Survey on Individual Differences in Visualization
EEUROVIS 2020S. Oeltze-Jafra, N. Smit, and B. Wang(Guest Editors)
Volume 39 ( ), Number 3STAR – State of The Art Report
Survey on Individual Differences in Visualization
Zhengliang Liu , R. Jordan Crouser , and Alvitta Ottley Washington University in St. Louis, USA Smith College, USA
Abstract
Developments in data visualization research have enabled visualization systems to achieve great general usability and ap-plication across a variety of domains. These advancements have improved not only people’s understanding of data, but alsothe general understanding of people themselves, and how they interact with visualization systems. In particular, researchershave gradually come to recognize the deficiency of having one-size-fits-all visualization interfaces, as well as the significanceof individual differences in the use of data visualization systems. Unfortunately, the absence of comprehensive surveys of theexisting literature impedes the development of this research. In this paper, we review the research perspectives, as well as thepersonality traits and cognitive abilities, visualizations, tasks, and measures investigated in the existing literature. We aim toprovide a detailed summary of existing scholarship, produce evidence-based reviews, and spur future inquiry.
1. Introduction
The term individual differences refers to individuals’ “traits orstable tendencies to respond to certain classes of stimuli or sit-uations in predictable ways” [DW96]. Much of the literature onindividual differences has roots in psychology. Psychological re-search has demonstrated that people with distinct personality typesand various cognitive abilities exhibit observable differences intask-solving and behavioral patterns [WB00,Ajz05]. Studies datingback to the late 1920s began by investigating variations in work-place performance [Hul28]. Throughout the intervening century,these findings have been extended to investigate individual char-acteristics that may predict performance under various conditions.In the past few decades, the computational sciences have begunto recognize the role individual differences might play in shapinginteraction in human-machine systems. For example, Benyon andMurray observed a relationship between spatial ability (a metricthat measures a personâ ˘A ´Zs ability to mentally represent and ma-nipulate two- or three-dimensional objects) and task performanceand preferences when using common interaction paradigms suchas menus and the command line [BM93]. Nov et al. [NALB13]found that extraversion (one’s tendency to engage with the exter-nal world) and neuroticism (a measure of emotional stability) hadeffects on users’ contributions to online discussions, and suggestedadaptations to certain visual cues to cater to different personalitytypes. Gajos and Chauncey [GC17] observed that introverted peo-ple were more likely to use adaptive features in user interfaces ascompared to extraverts . Orji et al. [ONDM17] showed that con-scientious participants (a measure of carefulness or diligence) re-sponded well to persuasive strategies such as self-monitoring andfeedback in gamified systems. These studies are just a small sampleof a large body of work documenting the influence of personalityand cognitive ability on interactions with computer interfaces. Formore detailed surveys of the literature, see [AA91, Poc91, DW96]. There is a growing interest in extending these findings to the fieldof data visualization [Yi12, ZOC ∗ ∗ ∗ ∗
14, OYC15] anduser satisfaction [Kob04]. Building on these findings, others havebegun to examine how we might leverage cognitive traits for ap-plications such as user modeling [BOZ ∗
14, OYC15] and adaptiveinterfaces [LTC19].In some circumstances, the interaction between individual differ-ences and visualization use may have critical impact on importantdecision-making processes. Ottley et al. [OPH ∗
15] investigated theimpact of visualization on medical decision-making, and found thatapproximately 50% of the studied population were unsupported bycommonly-used visualization tools when making decisions abouttheir medical treatment. Specifically, their study showed that vi-sual aides tended to be most beneficial for people with high spa-tial ability , while those with low spatial ability had difficulty inter-preting and analyzing the underlying medical data when they werepresented with visual representations. Another study by Conatiand Maclaren [CM08] found that participants with high perceptualspeed were less accurate in computing derived values when usingradar graphs instead of heatmapped tables for data analysis. A se-ries of studies have shown that locus of control (a measure of per-ceived control over external events) mediates search performanceon hierarchical visualizations [GJF10, GF12, ZCY ∗
11, ZOC ∗ c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization Unlike in human-computer interaction, to date there exists nocomprehensive report that surveys the relevant literature on therole of individual differences in the data visualization domain. Thismakes it difficult to understand the scope of existing research onindividual differences in this discipline, as there is no central re-source researchers can consult to learn what individual differences,visualizations and tasks have been studied, and whether the resultsof those studies have been independently replicated. More impor-tantly, there is limited information about how each existing studycontributes to the ultimate goal of designing flexible data visualiza-tion tools that better support individual users.In this STAR, we aim to produce a comprehensive survey thatreviews the literature relevant to this topic. We identify and tax-onomize existing scholarship to provide a complete picture of thecurrent state of research, and identify possible avenues for inves-tigation that builds upon this existing body of work. We begin bydescribing the scope of our review and methodology. We then pro-ceed to a detailed analysis of the findings of this body of work.Finally, we reflect on our review to discuss core topics and oppor-tunities for future development in this emerging area.
2. Existing Perspectives on Individual Differences inVisualization
The sampling of scholarly work in the previous section demon-strates the wide variety of individual differences that may be rel-evant to the visualization community. Pioneering work by Pecket al. [PYH ∗
12] proposed the Individual Cognitive Differences(
ICD ) model which classified the space of individual differencesinto three dimensions (see Figure 1): • Cognitive traits are the relatively stable characteristics of an in-dividual that include features of a person’s personality alongsidetheir cognitive abilities, such as perceptual speed , spatial ability ,and visual memory . • Cognitive states are temporary mental states such as cognitiveload and emotion. They are, by definition, transient and relatedto recent stimuli and the surrounding environment. • Experience is the long-term construction of knowledge throughexposure to real-world stimuli.
Bias describes the predisposi-tions one has such that one behaves in certain ways when per-forming certain tasks. Together, experience and bias represent adimension that describes the accumulation of experiences thatinfluence behavior when encountering a familiar scenario.Efforts to systematically evaluate visualization literacy (a measureof visualization experience for non-experts) [ARC ∗
17, BRBF14,BMBH16, DJS ∗
09, LKK16] postdate the
ICD model, but this canbe viewed as a specific domain of familiarity .In this STAR, we restrict the scope of our survey to focus only onthe invariant characteristics that distinguish one person fromanother . Unlike cognitive states and measures of experience , the cognitive traits covered in this survey are believed to be stablethroughout adulthood. This makes it tractable to reason about howthe community can begin to incorporate individual difference intodesign and evaluation pipelines. Our goal is to advocate for theadvancement of individual difference research in the visualizationdiscipline by highlighting the pioneering work in this domain. Figure 1:
The ICD model from Peck et al. [PYH ∗
12] categorizesindividual differences into three orthogonal dimensions: cognitivetraits, cognitive states, and experience/bias. In this STAR, we focusexclusively on cognitive trails.
3. Survey Scope and Methodology
This STAR report surveys the ongoing research that studies the im-pact of individual differences on the use of data visualizations. Thecandidate papers are obtained via three methods. First, we obtainthe main corpus by reviewing all the papers published on leadingconferences and journals in Visualization and HCI, including In-foVis, VAST, SciVis, EuroVis, TVCG, CHI and IUI from 2008 to2020. For this initial set of seed papers, we limit the scope to pa-pers published in Computer Science venues (e.g., we do not col-lect publications from PubMed, a search engine for biomedical andlife science references). Second, we search Google Scholar, ACMDigital Library, and IEEE Digital libraries with keywords such as individual differences , personality , cognitive ability and filter thereturned results to retrieve only data visualization publications. Wealso web-scraped ACM Digital Library and IEEE Digital Librariesto programmatically aid the process. Finally, we followed the cita-tions of the candidate papers obtained in the first two methods toexpand the scope of our seed paper to include relevant publicationsthat which were not published in computer science venues or werepublished before 2008. For all candidate papers we have collected,we manually review the title, abstract, introduction and conclusionsto determine whether they are within our proposed scope. If indoubt, we also review the main content of a paper to determine itsinclusion or exclusion. Finally, we removed duplicates manuscriptsthat studies that the same dataset. For example, [ZOC ∗ ∗ We compiled a corpus of relevant literature and organized the priorwork based on the types of individual differences , the visualizations used in the studies, and experimental designs such as the tasks and measures used in the experiments. During the first round of coding,a single author thoroughly read all papers to create an initial set c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization Table 1:
The 29 key articles we reviewed. The filled boxes indicate the traits ( ), visualizations ( ), tasks ( ), and measures ( ) thatwere in the manuscripts. We use to indicate traits that were evaluated, but no measurable effect was reported under the studied conditions.An interactive version of the table is available at https://washuvis.github.io/personalitySTAR . Traits Visualizations Tasks Measures
Five-Factor Model (cid:122) (cid:125)(cid:124) (cid:123) S a m p l e S i ze E x t r a v e r s i on N e u r o ti c i s m O p e nn e ss C on s c i e n ti ou s n e ss A g r eea b l e n e ss L o c u s o f C on t r o l N ee d f o r C ogn iti on S p a ti a l A b ilit y P e r ce p t u a l S p ee d V i s u a l/ S p a ti a l M e m o r y V i s u a l W o r k i ng M e m o r y V e r b a l W o r k i ng M e m o r y A ss o c i a ti v e M e m o r y S i m p l e V i s u a li za ti on S t a ti s ti ca l G r a ph s H i gh - d i m e n s i on a l S p a ti a l S ea r c h / R e t r i e v e V a l u e F i nd E x t r e m u m C o m pu t e D e r i v e d V a l u e S o r t I n f e r e n ce D r a w i ng / S p a ti a l R ea s on i ng D ec i s i on R ea d i ng C o m p r e h e n s i on S p ee d A cc u r ac y E y e T r ac k i ng M ou s e D a t a S ub j ec ti v e F ee db ac k I n s i gh t Vicente et al. (1987) [VHW87] 30Chen & Czerwinski (1997) [CC97] 11Chen (2000) [Che00] 10Velez et al. (2005) [VST05] 56Cohen & Hegarty (2007) [CH07] 30Conati & Maclaren (2008) [CM08] 45Ziemkiewicz & Kosara (2009) [ZK09] 63Green et al. (2010) [GJF10] 50Green & Fisher (2010) [GF10] 106Toker et al. (2012) [TCCH12] 35Ziemkiewicz et al. (2012) [ZOC ∗ ∗ ∗ ∗
14] 118Lallé et al. (2015) [LTCC15] 95Ottley et al. (2015) [OYC15] 108Ottley et al. (2015) [OCZC15] 300Vanderplas & Hofmann (2015) [VH15] 38Ottley et al. (2015) [OPH ∗
15] 377Conati et al. (2017) [CLRT17] 166Lallé & Conati (2019) [LC19] 46Millecamp at al. (2019) [MHCV19] 105Toker at al. (2019) [TCC19] 56Sheidin at al. (2020) [SLC ∗
20] 40 c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization of keywords. A second author then re-read the papers and addedor consolidated the keywords when there were gaps or redundan-cies. For the final round of coding, two researchers who were notinvolved in the previous two rounds validated the coding tags andpopulated Table 1. The three coding rounds were not completelyindependent, therefore, we do not measure coding coherence.
4. Overview of Paper
The proposed taxonomy of the publications consists of four dimen-sions: (1) the
Individual differences/traits studied; (2) the typesof visualization used; (3) the tasks involved in the associated ex-periment; and (4) the measures (or dependent variables) that wereevaluated. Table 2 summarizes the 11 primary traits observed in theliterature, which are used to organize the remainder of this paper.We classified each paper based on the dimensions in our pro-posed taxonomy. A natural way to accomplish this is to assign eachpaper one or more tags for each of the four dimensions. For exam-ple, the earliest paper in our collection by Vicente et al. [VHW87]explored how a series of traits might impact speed and navigationfor hierarchical search. Thus, the tags were
Spatial Ability, Per-ceptual Speed, Visual Working Memory, Networks, Search/RetrieveValue, Speed, Other Qualitative . Using these tags, we are able tosystematically analyze each paper following our taxonomy as aguide, distinguishing between
Personality Traits and
CognitiveAbilities . We visualize the tagging results in Table 1.
5. Personality Traits
Personality traits are the individual differences in thinking and be-having characteristics [All37]. The literature contains numerous ex-amples of personality traits that interact with visualization use. Forinstance, researchers have uncovered that locus of control , a mea-sure of perceived control, is a key factor that correlates with speed,accuracy and strategy [GJF10, GF10, BOZ ∗
14, OCZC12, OYC15,ZCY ∗ ∗ Five-Factor Model or Locus of Control . This is not surprising becausepsychologists have concluded that most personality traits are sub-sumed by the
Five-Factor Model [O‘C02].
Locus of Control hasalso been studied extensively in various domains [WL17].It is important to note that researchers commonly construct hy-potheses about the performance of individuals with different per-sonal characteristics based on theories and studies established inpsychology. We find that, in many cases, researchers will assumea trait to be advantageous to problem-solving with visualizations ifthis trait has been shown to be conducive to either problem-solving,decision-making, socioeconomic advancement or educational at-tainment, etc. For example, extraversion was hypothesized to behelpful in performing visual-related tasks [GF10]. However, per-sonality constructs are complex and interrelated, and we observeseveral cases in which the results are contrary to expectation.
Table 2:
Definitions of the cognitive traits that are common in the visualization literature.
Cognitive Traits
Extraversion
The tendency to engage with the external world.
Neuroticism
The tendency to experience negative emotions such as stress, depression or anger.
Openness to Experience
The propensity to seek, appreciate, understand and use information.
Agreeableness
The tendency to consider the harmony among a group of individuals. F i v e - F ac t o r M od e l [ G o l ] Conscientiousness
The propensity to control one’s impulse and display self-discipline.
Locus of Control [Rot66, Rot75, Rot90] The extent to which a person believes the external world is influenced by their own actions, and/orwhether they have control over the outcome of events occurring around them. P E R S ONA L I T Y T R A I T S Need for Cognition [CP82] The tendency to engage in and enjoy activities that involve thinking.
Spatial Ability [RS13] The ability to generate, understand, reason and memorize spatial relations among objects.
Perceptual Speed [EDH76] The rate at which an individual is able to make accurate visual comparisons between objects.
Visual / Spatial Memory [Spe63] The capacity to remember the appearance, configuration, location, and/or orientation of an object.
Working Memory [Bad92] The capacity to store information for immediate use. C OGN I T I V E A B I L I T I E S Associative Memory [Car74] The ability to recall relationships between two unrelated items. c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization Table 3:
The summary findings from Green and Fisher [GF10]
Completion Times Errors InsightsLocus of Control internal locusfaster times none external locusmore insights
Extraversion more extravertedfaster times none less extravertedmore insights
Neuroticism more neuroticfaster times none less neuroticmore insights
The five dimensions of the
Five-Factor Model (see [Gol93]) – ex-traversion , neuroticism (also referred to as emotional stability ), openness to experience , conscientiousness and agreeableness – arefrequently studied personalities among the surveyed publications(e.g., [ZK09, GJF10, GF10, ZCY ∗
11, ZOC ∗ ∗ Five-Factor Model . Some common survey instru-ments of the
Five-Factor Model include: Donnellan et al.’s Mini-IPIP [DOBL06] or De Young et al.’s 10 Big-Five Aspects [DQP07],and Johnson’s 120-question IPIP NEO-PI-R [Joh14].
Extraversion is defined as the tendency of an individual to engagewith the external world. Extraverts are more assertive and havestronger desire for social attention, compared to the more quiet andreserved introverts [WR17]. Extraverts have been found to achievehigher socio-economic status than introverts [Gen14]. Some studiesindicate a correlation between high level of extraversion and higheracademic achievements [CP13], while others have found contradic-tory results [HHL11]. The studies that find a negative correlationbetween extraversion and academic achievement hypothesize thatextraverts get distracted more easily than introverts [HHL11].
Extraversion in Visualization
A similar contradiction exists in the data visualization domain.Green et al. [GJF10] studied how extraversion (among others fac-tors) impact hierarchical search. Their initial studies found no cor-relation between extraversion and task performance. A follow upstudy with a larger sample size (106 versus 50 in their earlierstudy), however, revealed that extraverted participants completedsearch tasks more quickly [GF10]. In contrast, introverted subjectsattained more insights from the data [GF10]. Further investigationsby Ziemkiewicz et al. [ZOC ∗ extraversion and hierarchical search. Their resultsshowed no measurable effect on interaction time, but they foundthat extraversion impacted participants’ error rates. In particular,intraverted participants were more accurate in answering the ques-tions posed by the tasks.Altogether, the researchers found that extraverts and intro-verts exhibited different problem-solving approaches. The dif-ference in problem-solving approach was a likely explanationto the various reported results of the three studies. Specifically,Ziemkiewicz et al. [ZOC ∗ ∗ insights (seeTable 3). Insight in their study was defined as anything unexpectedor novel learned by the participants while completing the tasks. Re-searchers speculate that the extra time taken by introverts may bevery helpful in using data visualizations to solve problems, espe-cially for unfamiliar visualizations and datasets [ZOC ∗ Neuroticism is defined as the tendency to experience negative emo-tions such as stress, depression or anger. [JRSO14]. High neu-roticism is correlated with introversion [Uzi06] and low problem-solving skills [CRE ∗ neuroticism solved problems much faster than those witheither high or low levels of neuroticism [Far66]. Studies in the vi-sualization community, however, contradict this finding. Neuroticism in Visualization
Green and Fisher [GF10] found that more neurotic participantscompleted procedural tasks faster (see their summary findings inTable 3). Ziemkiewicz et al. [ZOC ∗ • More neurotic individuals are more attentive to tasks [IMB04],which is especially helpful when dealing with unfamiliar visual-izations and data [GF10, ZOC ∗ • More neurotic individuals are more likely to feel in control andmanipulate interfaces better, similar to those with internal
Locusof Control [GF10]. • More neurotic individuals might put more pressure on them-selves to perform the tasks well [ZOC ∗ • Since high levels of neuroticism are related to low emotion sta-bility , Ziemkiewicz et al. [ZOC ∗ • Less neurotic participants were either unwilling or less capableof adapting to the more unfamiliar, container style layouts andso performed poorly with those visualizations [ZOC ∗ Openness to experience (or " openness ") is defined as one’s propen-sity to seek, appreciate, understand and use information [DGP12].Being open to experience can be associated with being open-minded and curious. Psychologists have found that open to experi-ence is positively related to better academic achievement [HHL11,HVRT12]. A large-scale review by Jensen [Jen15] suggests thatsuch correlations have been found in many studies. Some scientists c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization Figure 2:
Visualization layouts used in Ziemkiewicz et al.’s studies on the influence of the five-factor model and locus of control on hierar-chical search tasks, each displaying the same data [ZCY ∗
11, ZOC ∗ also believe that openness is beneficial for professional develop-ment [NZ15] and that people who score high on openness tend tohave higher intellectual ability [AH97,GSL04] and better problem-solving skills [MSDL15]. Openness in Visualization
Researchers in the data visualization domain commonly constructhypotheses based on theories and prior results in psychology.However, openness is largely under-explored by visualization re-searchers. A single study by Ziemkiewicz and Kosara [ZK09]found that participants who scored high on openness had easiertime overcoming conflicting visual and verbal metaphors whensolving problems related to hierarchical visualizations. Brown etal. [BOZ ∗
14] found no measurable impact of openness on visualsearch strategies.
Conscientiousness is defined as the propensity to control one’simpulse and display self-discipline. [RJF ∗ ∗ ∗ ∗ conscientiousness scale is connected to being unreliable andlack of focus [MA03]. Overall, many researchers believe that high conscientiousness is related to career success [SCJME09, Tou12]and better problem-solving skills [DMOGP11]. Conscientiousness in Visualization
In the data visualization domain, however, conscientiousness hasnot been well-studied. In fact, to the best of our knowledge, thereis no recent publication that investigates how conscientiousness af-fects the use of visualizations. A few studies [ZK09,BOZ ∗
14] mea-sured conscientiousness alongside the other five-factor traits, butfound no significant impact.
Agreeableness measures a person’s tendency to consider the har-mony among a group of individuals [RC03]. Conversely, disagree-ableness/ low agreeableness is associated with prioritizing one’sself-interest.
Agreeableness is considered to be a beneficial traitfor performing collaborative tasks in teams [DGSO06, PVTRR06].Scoring low in agreeableness , however, can also be potentially ad-vantageous because some researchers have found that low agree-ableness is associated with creativity [KLWG13]. Also, there arecontradictory opinions on whether agreeableness is positively re-lated to academic achievement [LSLG03, HHL11] or not [Dis03].
Agreeableness in Visualization
As with to conscientiousness , agreeableness has yet to be studiedin-depth by visualization researchers although it has been mea-sured as part of the Five-Factor Model in a small number ofstudies [ZK09, BOZ ∗ ∗
14] found no effect on search tasks.
Locus of control measures the extent a person feels in controlof or controlled by external forces [Rot66, Rot75, Rot90]. Indi-viduals fall on a continuous spectrum, with one end being in-ternal locus of control ( Internals ) and the other end being ex-ternal locus of control ( Externals ). The Internal-External Locusof Control Inventory is a popular measure to evaluate an indi-vidual’s locus of control [Rot66]. Low scores are associated withinternal locus of control and high scores are associated with ex-ternal locus of control . According to Rotter [Rot90], individualswho exhibit internal locus of control believe that they have con-trol over their own actions, the actions’ outcomes and the environ-ment around them. In contrast, those who exhibit external locus ofcontrol tend to attribute outcomes to external forces.
Internals tendto be more confident [Hei10] and optimistic [BH15] than
Exter-nals . Researchers also believe that internal locus of control is asso-ciated with academic achievements [FC83, GBPM06]) and strongproblem-solving skills [MR93, OS15].
Green et al.’s experiments [GJF10, GF10] were among the firstto study the relationship between locus of control and userperformance with visualization-related tasks. They conducted astudy [GF10] to investigate the relationship between locus of con-trol and search performance across two hierarchical visualizationdesigns. They found that
Internals were significantly faster than
Externals when performing procedural tasks (search tasks to locateitems). However, locus of control had no significant impact on ac-curacy.
Externals , however, reported more insights than
Internals .Ziemkiewicz et al. [ZCY ∗
11] extended Green et al.’swork [GJF10, GF10] to further investigate how locus of control af-fects visualization use. They hypothesized that layout (defined asthe spatial representation and arrangement of visual marks in a vi-sualization [ZCY ∗ locus of control and visualization usage. They further hy-pothesized that Internals would have difficulties using visualization c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization Figure 3:
Visualization of different search pattern observed in Ottley et al.’s study, grouped by locus of control (external vs. internal) aswell as visual layouts [OYC15]. The thickness of the each line between every two nodes is proportional to the number of participants whoexplored that path. that were more "contained", while
Externals would be able to ad-just to various visual layouts. To test their hypotheses, Ziemkiewiczet al. [ZCY ∗
11] designed four visualizations that differed only inlayout. They designed and tested a set of visualizations that gradu-ally transitions from an indented list layout to a containment layout,while keeping constant the interaction mechanisms (e.g., zoomingv.s. scrolling), color encoding, and fonts. Figure 2 shows the fourvisual metaphors used by Ziemkiewicz et al. [ZCY ∗ ∗ ∗ Externals were faster and more accurate than
Inter-nals . The performance differences were especially pronounced inthe cases where participants used more "contained" visualizations(see the 3rd and 4th visual layouts in Figure 2). One interesting re-sult was that
Internals were significantly slower than
Externals incompleting inferential tasks [CWCO19] (such as comparing twoitems/objects found in the visualization), although
Internals and
Externals completed procedural tasks at approximately the samespeed. Ziemkiewicz et al. [ZCY ∗
11] speculated that
Externals werebetter than
Internals at adapting their thinking to external represen-tations (such as the layout of a visualization) since they were moreinclined to rely on external conditions rather their own internal rep-resentations and processes.Although locus of control is believed to be relatively stablethroughout adulthood, psychologists have found that it is possibleto temporarily influence a person’s locus of control score [JGPC92,FJ96]. Some researchers see this as an opportunity to resolve designchallenges. Further investigations by Ottley et al. [OCZC15] repli-cated Ziemkiewicz et al.’s [ZCY ∗
11] experiment design to studywhether changes in locus of control can predictably influence per-formance. The priming method used in their study was based onFisher and Johnson’s technique [FJ96]. This technique works byasking a person to recall examples of times when they feel eitherin control of (priming
Externals to be more internal ) or out ofcontrol of (priming
Internals to be more external ) the situations.The results of Ottley et al. showed that priming was largely ef-fective [OCZC15]. For example, when
Internals were primed tobe more external , they exhibited performance measures similar toparticipants grouped as
Average by Ziemkiewicz et al. [ZCY ∗ Average participants who were primed to be more inter- nal produced performance measures similar to the
Internals of thecontrol group. The only exception was
Average primed to be exter-nal . Their behaviors differed from the control group.In addition to these performance differences, researchers believethat it is also possible for locus of control to affect behaviouralpatterns [OYC15]. To investigate this, Ottley et al. [OYC15] an-alyzed the strategies employed by
Externals and
Internals withtwo different hierarchical visualizations (indented trees and den-drograms). For indented trees,
Externals followed the top-down de-sign of the indented tree and adopted a strategy similar to depth-first search, while
Internals followed a strategy that somewhatresembled breadth-first search. For dendrograms,
Externals weremore sporadic when they navigate the visualization, while
Inter-nals pursued a combined depth-first search and breadth-first searchstrategy. Figure 3 shows the various strategies observed (note thatthe thickness of a route is proportional to the number of partici-pants observed to follow that path). The results showed that
Exter-nals performed better (they found the targets faster) with indentedtrees, while
Internal were superior with the dendrogram. Similarly,Brown et al. [BOZ ∗
14] found that
Internals and
Externals applieddifferent searching strategies when performing a visual search task.
Cohen et al. first described need for cognition in 1955 as “a needto structure relevant situations in meaningful, integrated ways.It is a need to understand and make reasonable the experientialworld.” [CSW55]. In more recent conceptualization, the term hascome to mean a “chronic tendency to engage in and enjoy effortfulactivities” [CPFJ96], such as reading and solving puzzles.According to Cacioppo and Petty’s characterization of this con-cept [CP82], individuals with high need for cognition are morelikely to make sense of their world by seeking, acquiring, and re-flecting on information. In contrast, those with low need for cogni-tion are more likely to rely on others (e.g., experts and famous peo-ple), heuristics, or social comparisons to make meaning of events,relationships, and other stimuli. One common tool for assessingneed for cognition is a 34-item instrument developed by Cacioppoand Petty [CP82], which scores participants along a continuumfrom low to high need for cognition . A later version condensed c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization the number of items to 18, with no appreciable loss of discrimina-tory power [CPFK84], and this short form is the most common toolused to measure need for cognition in visualization-related studies,e.g. [CM08, MHCV19, TCC19].Several studies have sought to evaluate the correlation between need for cognition and other measures of individual difference(see [CPFJ96] for a complete survey). Amabile at al. observed asignificant positive correlation between need for cognition and in-trinsic motivation, as well as a corresponding negative correlationwith extrinsic motivation [AHHT94]. Fletcher at. al found that peo-ple with a higher need for cognition tended to have a significantlymore internal locus of control [FDF ∗ need for cognition and the conscientiousness and openness dimensions of the Five-Factor Model [CPFJ96], but as of this writing, this link has notbeen experimentally validated. In an early investigation of the effect of need for cognition in visu-alization, Conati and Maclaren [CM08] conducted a study to evalu-ate the efficacy of various individual differences (including need forcognition ) in predicting the relative effectiveness of a radar graphand a heatmap for various tasks. They found that in conjunctionwith other measures, need for cognition had a positive relationshipwith accuracy in sorting tasks using the heatmap . They also foundthat this relationship was not present in trials utilizing the radarchart. However, the authors note that while they did observe a sta-tistically significant effect, the models explain only a small propor-tion of the overall variance, suggesting that the effects of need forcognition are likely moderated by other, yet unobserved features.Millecamp et al. found that need for cognition plays a role in aperson’s response to visual explanation of recommendations in amusic recommender system [MHCV19]. Using a custom recom-mendation interface built on top of Spotify (see Fig. 4), the studyvaried whether or not participants interacted with a baseline systemor with an augmented version including both bar chart and scatter-plot views containing information regarding why a selected songwas recommended. They observed a statistically significant inter-action effect between need for cognition and the participants’ sub-jective ratings of confidence. Specifically, there was a modest in-crease in confidence for participants with low need for cognition inthe visual explanation condition compared with the baseline, and amodest decreased in confidence for participants with high need forcognition in the visual explanation condition compared with thebaseline.Toker et al. [TCC19] observed that need for cognition had a sig-nificant positive effect on participants’ accuracy when performingrecall tasks with a bar chart as a component of a Magazine-StyleNarrative Visualization, but that there was no statistically signifi-cant relationship to speed. This may at first seem counterintuitive:one would expect that participants with higher need for cognition would be able to perform more quickly, and that their commitmentto synthesizing all available information would improve their ac-curacy. Upon closer inspection, however, we observe that in thisstudy, the term speed refers to the total time spent interacting with
Figure 4:
The music recommender interface from Millecamp et.al’s 2019 study on the effects of need for cognition on participants’response to visual explanation of recommendations. The interfacesfor the control condition differed from the stimulus condition onlyby the omission of the two highlighted regions, which provide ex-planations about why a song was recommended. the visualization. When this meaning is applied, the positive re-lationship between need for cognition and time spent interactingwith a visualization are in line with observations made in non-visualization contexts: because people with higher need for cog-nition are predisposed to engage in sensemaking behavior, it makessense that they would spend more time trying to understand thevisualization before moving on to the subsequent task. However,these findings were inconsistent with a followup study by the sameauthors [TCC19], wherein they reported no significant relationshipto time on task but did observe a relationship with accuracy. Addi-tionally, this latter study included an analysis of eye-tracking data,but found no significant relationship [TCC19]. These conflictingresults suggest that more investigation is needed into the role of need for cognition in visualization use.
6. Cognitive Abilities
Cognitive abilities refer to mental capabilities in problem solv-ing and reasoning (including visual reasoning ) [IB15]. The datavisualization community has extrapolated the effects of cogni-tive abilities on the users’ performances and experiences with vi-sualizations from foundational research in psychology. We findliterature related to spatial ability [CC97, Che00, VST05, ZK09,FTES13, OPH ∗
15, VH15], perceptual speed [CM08, TCCH12,TCSC13, SCC13, CCH ∗ ∗ visual workingmemory [CM08, DMBM09, APM ∗
11, TCCH12, SCC13, TCSC13,CCH ∗ verbal working memory [TCCH12, TCSC13,SCC13, CCH ∗ associative memory [Che00]. Spatial ability is broadly defined as the capacity to generate,understand, reason and memorize spatial relations among ob-jects [RS13]. Though there is no consensus on precisely whichmental abilities are encompassed by this general term, com-monly referenced components include spatial orientation, spa-tial location memory, targeting, spatial visualization, disembed-ding and spatial perception (for further detail on these concepts,please see [Kim00]). Individuals with high spatial ability tend c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization to excel in scientific and engineering fields [WLB09] and ex-hibit stronger problem-solving skills for various tasks [WHA ∗ ∗ spa-tial ability include the paper folding test [EDH76] and mental ro-tation test [VK78]. Given the importance of spatial ability in analytical contexts, therelationship between this construct and visualization use has gen-erated substantial interest in the visualization research community.Early work by Vicente et al. [VHW87] investigated how spatialability influenced interactions with computer-based visualizations.In this study, participants were asked to locate a piece of informa-tion in a hierarchical file system. The researchers found that spatialability was a significant predictor of completion time, and they con-cluded that spatial ability had a dramatic impact on performance.Later studies found the spatial ability ’s influence on visualizationuse and performance might not be as straightforward as one wouldexpect. In the information retrieval domain, Chen and Czerwin-ski [CC97] reported that spatial ability was positively correlatedwith recall, but negatively correlated with precision, and these find-ings were partially replicated in a follow up study [Che00].Most studies, however, have consistently reported that spatialability is positively correlated with performance in various visualtasks. For example, Velez et al. [VST05] found that participantswith higher spatial ability were faster and more accurate at iden-tifying real and computer-generated 3-D objects when given theobjects’ orthogonal projections from various perspectives. Cohenand Hegarty [CH07] asked participants to sketch the cross sectionof a computer generated 3-D object, and observed that individu-als with higher spatial ability generally performed better, and thatthese same participants were more likely to make use of supportinganimation.As in their investigation of openness , Ziemkiewicz andKosara [ZK09] observed that individuals with high spatial abil-ity were better equipped to overcome incompatible visual and ver-bal metaphors when navigating hierarchical data. In their study onthe efficacy of visualization and structured text in supporting med-ical decision-making, Ottley et al. [OPH ∗
15] reported that partici-pants with higher spatial ability were more accurate and faster thanthe group with low spatial ability , and were better able to makeuse of the more text+visualization representation of the data. Simi-lar performance advantages were reported in VanderPlas and Hof-mann’s [VH15] experiment with lineup tasks, and Conati and Ma-claren [CM08] found that spatial ability was positively correlatedwith better performance in characterizing distributions. However,Froese et al. [FTES13] found that people with low spatial abil-ity experienced significant performance gains after being trained inusing visualizations.These studies broadly suggest that spatial ability has a largelypositive relationship with performance when using visualizations.One possible explanation is that spatial ability might affect aparticipant’s strategy or usage pattern. For example, Vicente etal. [VHW87] found that individuals with low spatial ability fre-quently descended an incorrect path through the hierearchy, requir- ing them to backtrack. Chen and Czerwinski [CC97] observed thatparticipants with high spatial ability commonly combined detailedlocal moves with strategic jumps that exploited the global structureof the visualization, whereas those with low spatial ability tendedto remain at the local level.
Perceptual speed measures the rate at which an individual canscan and compare figures and symbols, as well as perform sim-ple visual perception tasks [EDH76]. Studies have demonstratedlinks between high perceptual speed and educational achieve-ment [Mel82], information retrieval [All92], and acquiring pro-gramming skills [Shu91]. Some commonly used tests for percep-tual speed are the
Identical Pictures Test [EDH76], the
Finding A’sTest [EDH76] and
Number Comparison Test [EDH76].
Vicente et al.’s [VHW87] pioneering study on spatial ability in-cluded perceptual speed as one of the candidate predictors of userperformance. However, they found no measurable effect of percep-tual speed on searching hierarchical file systems. More recent in-vestigations by Conati and Maclaren [CM08] found that percep-tual speed mediate tasks performance. For example, they found perceptual speed to be positively correlated with the accuracy of"computing derived values", a category of tasks defined by Amaret al. [AES05] that involves deriving an aggregate number fromgraphical data. Overall, the found that participants with low per-ceptual speed did better than those with high perceptual speed with radar graph, while the opposite was true for heatmapped ta-bles [Wil04] (see Figure 5). Toker et al. [TCCH12] also foundthat individuals with high perceptual speed completed tasks fasterwith both radar and bar graphs. Similar results were reportedby [CCH ∗ ∗ perceptual speed led to higherlearning rate (measured by the change in task completion time oraccuracy over time) [LTCC15]. Toker et al. [TCC19] found thatindividuals with low perceptual speed had difficulties remember-ing legend details and axis labels. Further studies by Toker etal. [TCSC13], Steichen et al. [SCC13] and Conati et al. [CLRT17]all showed that it was possible to infer a user’s perceptual speed dynamically based on eye-tracking data. Figure 5:
The radar graph (A) and heatmapped tables (B) used byConati and Maclaren in their exploration of the relationship be-tween perceptual speed and task performance [CM08]. c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization Visual / spatial memory measures the short-term ability to re-member the configuration, location, and orientation of an ob-ject [Spe63], and is commonly measured using Eckstrom et al.’sShape Memory Test (MV-1) [EDH76] or other similar instruments.The visuospatial nature of data visualization suggests an intuitivelink between an individual’s visual memory and their performanceusing visualization tools, and this intuition has led to an abundanceof studies investigating this relationship. However, the results ofthese investigations have been mixed.
Several of the studies described in Sections 6.1 and 6.2 also in-vestigated the role of visual memory. In Vicente et al.’s 1987study found no relationship between visual working memory andhow people navigate hierarchical file systems [VHW87]. Similarly,Chen’s 2000 study [Che00] ( see section 6.1) found no relation-ship between visual memory and search performance in a spatial-semantic virtual environment. Velez et al. [VST05] did observe astatistically-significant relationship between visual memory and ac-curacy in their projection task, but the influence was modest. Conatiand Maclaren [CM08] reported a similar relationship during filtertasks using the heatmapped table, but not the radar chart.Participants with low spatial memory in Lallé at al.’s study ofuser experience reported that they found the MetroQuest interfacesubstantially less useful [LCC17, LC19]. In a companion analysisto this study, Conati et al. [CLRT17] found that eye tracking datacould be used to accurately predict participants’ spatial memory ,suggesting that this feature is associated with distinct gaze patternsin visualization use.
Many of the studies that investigated perceptual speed also eval-uated working memory , a measure of an individual’s capacity fortemporarily storing and manipulating conscious perceptual andlinguistic information [Bad92, MS99]. This term was originallycoined in 1960 by Miller et al. in the context of their work on theoryof mind [MGP60], and is distinct from short-term memory in thatthe emphasis is on the active manipulation of information, ratherthan simple recall [Cow08].Daneman and Carpenter first observed a link between workingmemory and reading comprehension [DC80], and this relationshiphas been independently verified by many other studies [DM96]. Itappears to play a substantial role in academic achievement [SBF04,AA10], as well as in attention [FV09], though the latter finding hasbeen recently called into question following a more nuanced in-vestigation using eye tracking [MMWL14].
Working memory is ofparticular interest to visualization researchers because of its signif-icance in supporting reasoning [Voo97, Kla97, CHD03], decision-making [HJW02, HJW03, Brö03], and other cognitive processescritical to effective analysis [Dia13].
Two different forms of working memory are frequently assessedin the visualization literature.
Visual working memory is a mecha- nism by which visual information (including position, shape, color,and texture) is retained between eye fixations [LV97]. This enablescognitive actions such as change detection [LV13].
Verbal work-ing memory is responsible for temporarily storing and manipulat-ing language-related information, including both words and numer-ical values [vDM16]. This enables actions such as rememberinga telephone number long enough to dial it [MD16]. A commonlycited test for visual working memory is a set of change detectiontasks of colored squares developed by Edward K. Vogel and col-laborators ( [LV97, VWL01, FV09]). For measuring verbal work-ing memory , Operation-Word Span Test (OSPAN) [TE89] and theCorsi Test [Cor72] are found in the surveyed literature.Toker et al. [TCCH12] found a statistically-significant, diver-gent relationship between participants’ working memory and theirpreference ratings of bar charts and radar plots. Specifically, par-ticipants with higher visual working memory rated radar graphsas more preferable, and those with lower verbal working memory tended to rate bar graphs as easier to use. In follow-ups to this studyusing the same interfaces and tasks, Steichen et al. [SCC13,SCC14]found that eye tracking data could be used to accurately predictboth visual and verbal working memory . However, further analysisfound that only verbal working memory was statistically significantin the prediction of specific gaze behaviors [TCSC13].In their investigation of the effects of highlighting interventionson speed and accuracy on search and comparison tasks using barcharts, Carenini et al. [CCH ∗ visual and/or verbal working memory consistently underperformedon comparison tasks. This effect was absent for simple search tasks.Conati et al. observed similar relationships in their study of ValueCharts [CCH ∗ working memory usefulin predicting participants’ learning curve on this interface, charac-terized by the rate of change in response time over multiple trials.In their experiments on the MetroQuest system, Lallé atal. [LCC17] observed a relationship between visual working mem-ory and both user preference and gaze behavior. Specifically, par-ticipants with higher visual working memory tended to prefer chartsover maps, and correspondingly tended to have more fixations onthe chart areas. In these experiments, there was no relationship ob-served between working memory and willingness to utilize avail-able interface customization options [LC19]. As with spatial mem-ory , a deeper analysis of the gaze data from this experiment byConati et al. (2017) [CLRT17] demonstrated that gaze data can beused to accurately predict participants’ visual working memory .In Toker et al.’s experiments on Magazine Style Narrative Visu-alizations [TCC19], verbal working memory was observed to havean intuitive negative correlation with time on tasks. They did not,however, observe any statistically significant correlation with ac-curacy, understanding, or interest. While visual working memory was measured in participants of Millecamp et al.’s experiments in-volving a more music curation task, they also did not observe anystatistically significant relationship with this feature. c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization Associative memory refers to a person’s ability to recall a relation-ships between two unrelated items (for example, linking a nameand a face) [Car74]. It can be measured by MA-1 scores [EDH76].Some researchers believe that it is valuable to investigate the effectsof associative memory on user interaction with data visualizations,because good associative memory helps building mental maps ofvirtual environments or interfaces and can aid users in navigatingthe virtual spaces [Che00].
To the best of our knowledge, Chen [Che00] is the only publicationto investigate associative memory in the existing data visualizationliterature. In Chen’s study, participants used an interactive graphof published articles and were asked to retrieve as many papers aspossible for a given topic within 15 minutes. Chen found that as-sociative memory was positively correlated with people’s ability toretrieve the appropriate papers. Chen also reported the subjectivefeedback of users and found that those with good associative mem-ory were more likely to believe that the spatial interface was useful.
7. Discussion of Findings
Although our organization of the literature on individual differ-ences in visualization is intended to provide a broad overview ofexisting work in this area, we acknowledge that any post-hoc cat-egorization (such as the traits, visualizations, tasks, and measuresreported in this STAR) will not be exhaustive. Despite this fun-damental limitation, our taxonomy enabled several useful insightsregarding this body of work. Foremost among them were two im-portant takeaways:1. With very few exceptions (namely, conscientiousness and agreeableness ), there is evidence that nearly every cognitivetrait in Table 2 can impact visualization use . This body ofwork underscores that designing and evaluating tools to helppeople think is a complicated endeavor.2. Despite the breadth of cognitive traits under investigation, therehave been a relatively small number of studies which at timesyielded conflicting findings. Further investigations, includingreplication studies , are crucial to enriching our understandinghow individual differences impact visualization use, and to sub-sequently develop guidelines for the integration of this knowl-edge into the design of future systems.In the following sections, we expand upon these observations in thecontext of several different dimensions of our taxonomy. The impact of some individual differences are clear, having beenreplicated under multiple experimental conditions by two or moreindependent researchers. One such example is the consistentdemonstration that locus of control impacts speed and accuracyon hierarchical search tasks [GJF10, ZOC ∗ ∗ ∗ locus of control influences search strategy [OYC15].It is interesting to note that verbal working memory is the onlytrait that has reliably resulted in statistically significant findings. Verbal working memory is believed to affect the processing of ver-bal component of visualizations, such as labels, legends, descrip-tion of tasks, and texts [TCCH12, TCSC13, TCC19]. In particular,high verbal working memory users spend less time reading and pro-cessing various textual information in visualizations [SCC13]. Ananalysis of eye tracking data by Toker et al. [TCSC13] indicatedthat participants with low verbal working memory referred backto task question descriptions more frequently, and tended to scanbetween different parts of the screen more frequently than theirhigh verbal working memory counterparts. Another study foundthat verbal working memory was positively correlated with learn-ing rate [LTCC15]. Overall, studies have consistently reported aninverse correlation between task completion time and verbal work-ing memory , though we hesitate to generalize these findings to real-world scenarios. This correlation may be attributable to unintendedsituational effects of the design of traditional user studies, whichexplicitly require participants to process textual information whencompleting visualization-related tasks.Results are more ambiguous for most other traits that have beenstudied due to the lack of replication studies. For example, althougha series of manuscripts report that perceptual speed impacts visu-alization use (9 out of 29 papers report significant effects), theyinspected a range of visualization designs, tasks, and measures,making it challenging to uncover general patterns. A similar phe-nomenon exists for visual working memory. Although every paper
Figure 6:
The types and distribution of traits that were investigatedin the literature on individual differences in visualization use. c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization Figure 7:
The types and distribution of visualization designs ob-served in the literature on individual differences in visualizationuse. in this STAR used established psychometric batteries from the psy-chology field, inconsistency among the surveys used to assess traitsalso makes it difficult to compare findings between studies. For ex-ample, both Ottley et al. [OPH ∗
15] and Micallef et al. [MDF12]have investigated the impact spatial ability on Bayesian inferencewith visualization, but reported contradictory findings. Both studiesused the same paper folding task to assess spatial ability [EDH76],but differed in the application of the assessment instrument: Mi-callef et al. [MDF12] used 10 out of 20 questions in the scale, whileOttley et al. [OPH ∗
15] used all 20 questions. Such inconsistenciesunderscore both the importance of replication and the need to stan-dardize the instruments used to assess both individual differencesand task performance.Other traits remain underexplored despite promising initial find-ings. As mentioned in Section 6.5, Chen [Che00]’s singular studyon associative memory showed a positive correlation betweenthis trait and performance on a graph navigation tasks. Similarly,Ziemkiewicz and Kosara [ZK09] found that openness to experience predicted easier adjustment to disruptions in visualization interac-tion, an observation which has promising implications for visual-ization scenarios involving unfamiliar or novel designs. Other traitssuch as conscientiousness are also sparsely explored in the contextof visualization use, with only 2 of out 29 manuscripts inspectingthis trait. Both studies reported null results, though it is impossi-ble to draw comparisons between these studies due to their vastlydifferent experimental designs.
We observed five categories of visualization design in the sur-veyed literature: Simple Visualization, Statistical, Graphs, High-Dimensional, and Spatial.
Graphs were the most commonly testedvisualization in the individual differences literature, appearing in9 out of 29 surveyed papers. We observed substantial variance inthe choice of both encoding and aesthetic design. The research ex-ploring hierarchical visualization has largely focused on the impactof locus of control [GF12, OYC15, OCZC12, ZCY ∗
11, ZCY ∗ five factor model [GF12, ZCY ∗
11, ZCY ∗
11, ZK09]. Sev-eral studies also report that search and navigation with graphs andtrees is influenced by spatial ability [Che00,CC97,VHW87,ZK09].
Simple data visualizations were also relatively common in the literature (8 out of 29 papers surveyed). For example, Toker etal. [TCSC13] and Steichen et al. [SCC13] found that perceptualspeed , visual working memory , and verbal working memory caninfluence how people deploy attention to visual elements withingrouped bar charts. VanderPlas and Hofmann [VH15] used his-tograms and dotplots among other charts, and found that spatialability correlated with performance when identifying which plotwas “the most different” in a collection.VanderPlas and Hofmann [VH15] also included statistical plots such as boxplots, violin plots and QQ-plots, and found a simi-lar correlation between performance and spatial ability . Micallefet al. [MDF12] and Ottley et al. [OPH ∗
15] investigated statisti-cal plots such as icon arrays (also known as frequency grids) andEuler diagrams. As reported in Section 7.1, these studies reportedcontradictory results on whether or not spatial ability influencedperformance on Bayesian inference tasks.A series of studies investigated individual differences in the con-text of radar plots (e.g., [CM08], [SCC13], and [TCSC13]). A laterstudy by Sheidin et al. [SLC ∗
20] compared speed and accuracyacross a variety of tasks with different time series visualizations,including line charts, stream graphs, radar charts, and circle charts.They found a significant interaction between locus of control andspeed and accuracy in some task types, and observed that verbalworking memory also influenced completion times. Taken together,these findings suggest a correlation between perceptual speed , vi-sual working memory , and verbal working memory and visualiza-tion use.Studies of spatial ability in the context of spatial visualiza-tion universally reported significant effects [CC97, Che00, CH07,FTES13, VST05]. For example, Chen et al. [Che00] observed that spatial ability was correlated with graph search performance in vir-tual environments. Froese et al. [FTES13] demonstrated that train-ing programs for creating projections of 3D objects were most ben-eficial for participants with low spatial ability . Figure 8:
The types and distribution of tasks observed in the liter-ature on individual differences in visualization use. c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization We observed a wide variety of tasks in the literature, many ofwhich are based loosely on Amar et al.’s analytic task taxon-omy [AES05] (e.g., information retrieval/search , find extremum , compute derived values , and sort ). Search was the most commontype of task that we observed in the literature. Examples range fromfinding documents in a file structure [Che00, VHW87] or phylo-genic tree [GF10,OCZC15,ZOC ∗ ∗ spatial ability were more accurate in recalling the vi-sualization structure. Toker et al. [TCC19] asked subjects to use a“Magazine Style Narrative Visualization” that supplemented tex-tual documents and answer reading comprehension questions. Mil-lecamp et al. [MHCV19] asked participants to use a custom inter-face (Fig. 4) to create music playlists for different activities. The majority of existing studies assessed the effects of personalitytraits and cognitive abilities through traditional measures of perfor-mance such as speed (20 out of 29 papers) and accuracy (20 out of29 papers). With a few exceptions (e.g. [CM08]), high scores in thestudied cognitive abilities were correlated with better task perfor-mance. For example, higher levels of spatial ability correlated withbetter statistical reasoning [OPH ∗ perceptual speed predicted superior ability to find similarities and differences amongobjects [All94].In contrast, the results for personality traits are more nuanced,with effects that are moderated more significantly by visualiza-tion design and task. For example, Ottley et al. [OYC15] comparedsearch speed across two tree visualization designs: a dendrogramand an indented tree. They found that participants with external lo-cus of control were faster than their internal locus of control coun-terparts when performing search tasks with an indented tree visu-alization. However, they also observed the reverse when they ana-lyzed interaction times for the dendrogram, suggesting that neitherof the designs were suitable for both groups of users. Figure 9:
The types and distribution of measures observed in theliterature on individual differences in visualization use.
In addition to speed and accuracy, studies frequently so-licited subjective feedback to evaluate users’ experiences [GJF10,TCCH12, LCC17, LC19, TCC19]. For instance, Ziemkiewicz etal. [ZOC ∗ associative memory weremore likely to believe that the graph-based visualization interfacewas useful. Toker et al. [TCCH12] found that participants withhigher visual working memory preferred radar charts more thanthose whose visual working memory was lower.Eye-tracking has seen significant use in the surveyed literature(7 out of 29 papers). Beyond measures of speed and accuracy, eyemovements can provide important information about how differ-ent representations facilitate information processing. Eye gaze datahas been used to predict both task and visualization type [SCC13],as well as task complexity as defined by the study [SCC14], userperformance [SCC14, TCC19], and learning curve [LTCC15]. Inaddition, one study found that perceptual speed was positively cor-related with more efficient visual scanning behavior [TCSC13].Three of the surveyed studies used mouse interaction to ex-plore how individual differences impact visualization use. Vicenteet al. [VHW87] found that spatial ability correlated with scrollingbehavior with a hierarchical file structure. Ottley et al. [OYC15]showed that different locus of control groups exhibited distinct pat-terns of mouse movement when searching hierarchical visualiza-tions. Finally, Brown et al. [BOZ ∗
14] used machine learning topredict locus of control , extraversion , and neuroticism from fea-tures that were derived from mouse clicks and mouse movementsduring a visual search task.A single study [GF10] captured insights, a term which carriesseveral definitions in the visualization community depending onthe context [CZGR09]. In their study, Green et al. [GF10] defineinsights as “items or concepts learned or added to the userâ ˘A ´Zsknowledge base.” They found that external locus of control , intro-version , and low scores on the neuroticism scale mapped to moreinsights with their studied visualizations. More than half (18 out of 29) of the studies recruited local col-lege students (both undergraduates and graduate students), facultyor staff members as user study participants. 8 out of 29 studiesrecruited test subjects online, primarily from crowdsourcing plat-forms such as Amazon Mechanical Turk (a platform to recruit userstudy participants; see [PCI10] for more information about this ser-vice). The remaining studies did not report how they recruited par-ticipants.Conducting user studies with crowdsourced participants is a rel-atively recent phenomenon, and it is becoming increasingly pop-ular [ZOC ∗ experimenter bias [PCI10, MDF12]. However, collectingdata from crowdsourced experiments might raise concerns about c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization data quality [OYC15,OCZC15,MHCV19]. Some participants maynot take the experiments seriously and do something else whileparticipating in a study [ZOC ∗ ∗ ∗ agreeableness , lower extraversion and higher neuroticism ). Overall, it is advisable to keep these dif-ferences in mind when deciding whether an experiment should beconducted via crowdsourcing or not. In addition, more future re-search is needed to explore designs and techniques that improvethe ecological validity of lab studies, because students belong to anarrow and selected subset of the general population.
8. Opportunities for Future Research
Visualization users differ greatly in experiences, backgrounds, per-sonalities and cognitive abilities, yet visualizations, much like othersoftware products, continue to be designed for a single ideal user.It would be clearly impractical to design each visualization for anindividual user. However, knowledge of broad differences betweenuser groups could be used to guide design for specific domains andto suggest multiple analysis modes or customization options in asingle system. The body of work on individual differences in visu-alization provides a foundation for achieving this goal. However,successfully translating research to practice warrants more work. Itis currently difficult for ordinary developers, with no background invisualization or social science research, to identify potential issueswith their design choices. Perhaps the best-supported cognitive traitis color vision deficiency. There exist several designer tools for test-ing or verifying the color inclusiveness of a design [Cob, Cola] orfor selecting palettes that are colorblind safe [Colb]. A key futuredirection is to enable practitioners, with no individual differencesresearch background, to foresee the effect of their designs. Moreinvestigation is needed so that we can provide clear guidelines forresearch and practitioners, and success in the research agenda couldtransform how we evaluate and design visualizations for differentuser groups, tasks, and domains. There are many open questionsand challenges.
The research projects in this survey all used psychological sur-veys to estimate a person’s cognitive traits. In real-world scenar- ios, however, it may be unrealistic to expect users to be sub-jected to a deluge of forms. Discovering new and unobtrusivemethods to capture cognitive state , cognitive trait , and experi-ence/bias will ultimately drive research in individual cognitive dif-ferences. For example, Brown et al. [BOZ ∗
14] showed how wemight detect user attributes by analyzing their click stream data,and others have demonstrated similar successes using eye track-ing data [SCC13, CLRT17, TCC19]. In the broader visualizationcommunity, we have seen increased interest in developing algo-rithms to model users’ behavior and in investigating how we canuse these techniques to improve visualization tools (examples in-clude [DC17], [OGW19] and [WBFE17]).Although the research on user modeling and individual differ-ences have largely been separate, analyzing their intersection couldopen the doors for many exciting future work. For instance, ana-lyzing the portions of the data explored by the user can indicatea user’s expertise and biases [WBFE17]. Brown et al. [BOZ ∗ actions (e.g., pans and zooms) uncovereddifferences that were mediated by user’s locus of control scores andpersonality traits [BOZ ∗ visual attention via eye-gaze can reveal differences in people withvarying perceptual speed and visual working memory [SCC13].Therefore, to successfully infer individual traits, future work mustconsider a comprehensive set of encodings that include actions,data, and visual features. The ability to automatically infer per-sonality traits and individual characteristics will open many oppor-tunities for tailoring visualization systems to better suit the user.However, bridging the gap between visualization and personalitypsychology can raise serious privacy concerns. It is important to beaware of the potential ethical challenges ahead, and take socially-responsive steps to mitigate the effects. Task design is critical to the success of an evaluation [Mun09],and researchers have created taxonomies for the types of tasksand interactions that are feasible for a given visualization (forexample [AES05], [Shn96], [YaKSJ07], and [ZF98]). For futurework, it is essential to recognize that "exploration" as a taskcarries several different meanings. Recent work by Battle andHeer [BH19] distinguishes between bottom-up exploration andtop-down exploration. Bottom-up explorations "are driven in re-action to the data" [AZL ∗
18] or "may be triggered by salient vi-sual cues" [LH14]. This type of exploration is open-ended andthe user’s instincts largely drives the interactions. Top-down ex-plorations, on the other hand, are based on high-level goals or hy-potheses [BH19, GZ09, LH14].One shortcoming of the prior work that investigates how individ-ual traits impact exploration paths is they study only goal-driven,top-down exploration tasks. Because of this limitation, we knowonly the effect that individual traits have on interactions for top-down exploratory data analysis with short study duration. We needsystematic studies to investigate the correlation between task typesand patterns of interactions, and how individual traits may mediateobservations over time. One possibility for expanding the body ofliterature is to investigate the impact of personality for more open-ended visual analytics tasks. The VAST Challenge [CGW14], for c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization example, produces synthetic data annually and the challenges aredesigned to reflect real-world tasks under realistic conditions. Ad-ditionally, personal visualization [HTA ∗
14] can incorporate datafor use in a personal context. Future work can leverage these richdatasets to observe longitudinal top-down and bottom-up explo-ration processes and to uncover patterns in personality groups.
When we take a closer look at the previous results, much of the ob-served patterns of behavior can be explained by local versus globallevel precedence in processing information. For instance, whensearching a tree visualization, participants with an external locus ofcontrol ( Externals ) were more likely to perform a depth-first searchwhile participants with an internal locus of control ( Internals ) weremore likely to perform a breadth-first search [OYC15]. A depth-first search strategy suggests a local precedence information pro-cessing while a breadth-first search indicates attention to globalfeatures and their relationships. As a result, Ottley et al. [OYC15]demonstrated that
Externals were faster and more accurate withindented tree visualization. It is possible that the design encour-ages a local exploration. Similarly, when searching for Waldo, wefound that
Externals were more likely to explore at a lower zoomlevel, paying attention to local features, while
Internals tended toonly zoom in when they believed they had identified the target.This preference for attending to local versus global features sug-gests a pattern of behavior that may generalize across visualizationdesigns. Future work is needed to investigate the relationship be-tween individual traits and processing precedence across differentdesigns.
One important advantage of understanding individual users’ cog-nitive traits , and biases as a cohesive structure is that this opensup the possibility of developing adaptive, mixed-initiative visual-ization systems [TC05]. Principles for similar mixed-initiative sys-tems were proposed in the HCI community by Horvitz [Hor99].As noted by Thomas and Cook in
Illuminating the Path [TC05],an important direction in advancing visual analytics research is thedevelopment of an automated, computational system that can assista user in performing analytical tasks. However, most visualizationsystems today are designed in a one-size-fits-all fashion without theability to adapt to different users’ analytical needs into the design.Creating such mixed-initiative visualization systems is partic-ularly difficult as visualization are often designed to supportcomplex thought and decision-making. Still, there is some evi-dence that successful adaptive systems can significantly improvea user’s ability in performing various tasks. For example, Gotz andWen [GW09] proposed a behavior-driven visualization recommen-dation system that infers visual analytic tasks in real-time and sug-gests visualizations that might support the task better. Other workdemonstrated how we can detect and adapt to mitigate explorationbiases [GSC16, LDH ∗
19, WBFE17]. It is clear that adaptive sys-tems can offer new possibilities for visualization research and de-velopment [GW09], but additional work is necessary to understand how and when a system should adapt to a user’s needs. Specifi- cally, studies are needed in order to carefully map features of theuser unto the visual or interaction encodings of the system.
9. Conclusion
The community has made great strides in identifying characteristicsthat could impact performance on visualization systems. However,this work is still in its infancy, and uncovering the correlation be-tween traits, visual design, and tasks is only the first step. Whatis clear from the existing body of research is that a mismatch be-tween cognitive traits and visualization design can result in a gulfof evaluation or execution [ND86].
The gulf of execution describesthe difference between the user’s intentions and how well the de-sign supports their goals.
The gulf of evaluation is the differencebetween the system’s state and the user’s perceived state of thesystem. Vicente et al. [VHW87] acknowledges this challenge: “Al-though the assay and the isolation phases locate the locus of theindividual differences in specific task components and certain usercharacteristics, they do not provide the designer with enough infor-mation to predict whether or not a given accommodation schemewill be successful.”Egan and Gomez [EG85] proposed a methodology for reducingthe impact of individual differences for a given interface. They rec-ommended a three-phased strategy: assaying, isolating, and accom-modating individual differences. Assaying involves identifying keycharacteristics, and isolating requires understanding the interactionbetween the user characteristics and specific task components. Fi-nally, the accommodation phase calls for changing or eliminatingthe problematic tasks.The vast majority of the existing work are still in the assayingand isolating phase, and the community has yet to provide tech-niques for accommodating the broad variety of visualization users.We believe that this manuscript can serve as a central resource forpractitioners and researchers to learn about the landscape of re-search on individual differences in visualization use and the im-plications for design and evaluation. Moreover, we hope this studywill inspire future work that completes the understanding of in-dividual differences and visualization, and serve as a catalyst fornext-generation data visualization tools that better support individ-ual users.
Acknowledgments
We thank Jesse Huang’19 (Washington University in St. Louis) forhis help with data collection, and Ananda Montoly ’22 (Smith Col-lege) for her work validating the paper classification. This projectwas supported in part by: the Laboratory for Analytic Sciencesat North Carolina State University, The Boeing Company underaward 2018-BRT-PA-332, and the National Science Foundation un-der Grant No. 1755734. c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization References [AA91] A
YKIN
N. M., A
YKIN
T.: Individual differences in human-computer interaction.
Computers & industrial engineering 20 , 3 (1991),373–379. 1[AA10] A
LLOWAY
T. P., A
LLOWAY
R. G.: Investigating the predictiveroles of working memory and iq in academic attainment.
Journal ofexperimental child psychology 106 , 1 (2010), 20–29. 10[AES05] A
MAR
R., E
AGAN
J., S
TASKO
J.: Low-level components ofanalytic activity in information visualization. In
IEEE Symposium onInformation Visualization, 2005. INFOVIS 2005. (2005), IEEE, pp. 111–117. 9, 12, 14[AH97] A
CKERMAN
P. L., H
EGGESTAD
E. D.: Intelligence, personality,and interests: evidence for overlapping traits.
Psychological bulletin 121 ,2 (1997), 219. 5[AHHT94] A
MABILE
T. M., H
ILL
K. G., H
ENNESSEY
B. A., T
IGHE
E. M.: The work preference inventory: assessing intrinsic and extrinsicmotivational orientations.
Journal of personality and social psychology66 , 5 (1994), 950. 7[Ajz05] A
JZEN
I.:
Attitudes, personality, and behavior . McGraw-HillEducation (UK), 2005. 1[All37] A
LLPORT
G. W.: Personality: A psychological interpretation.Holt. 4[All92] A
LLEN
B.: Cognitive differences in end user searching of a cd-rom index. In
Proceedings of the 15th annual international ACM SI-GIR conference on Research and development in information retrieval (1992), pp. 298–309. 9[All94] A
LLEN
B.: Perceptual speed, learning and information retrievalperformance. In
SIGIRâ ˘A ´Z94 (1994), Springer, pp. 71–80. 13[APM ∗
11] A
NDERSON
E. W., P
OTTER
K. C., M
ATZEN
L. E., S
HEP - HERD
J. F., P
RESTON
G. A., S
ILVA
C. T.: A user study of visualizationeffectiveness using eeg and cognitive load.
Computer Graphics Forum30 , 3 (2011), 791–800. 8[ARC ∗
17] A
LPER
B., R
ICHE
N. H., C
HEVALIER
F., B OY J., S
EZGIN
M.: Visualization literacy at elementary school. In
Proceedings of the2017 CHI Conference on Human Factors in Computing Systems (2017),ACM, pp. 5485–5497. 2[AZL ∗
18] A
LSPAUGH
S., Z
OKAEI
N., L IU A., J IN C., H
EARST
M. A.:Futzing and moseying: Interviews with professional data analysts on ex-ploration practices.
IEEE transactions on visualization and computergraphics 25 , 1 (2018), 22–31. 14[Bad92] B
ADDELEY
A.: Working memory.
Science 255 , 5044 (1992),556–559. 4, 10[BH15] B
RYANT
F. B., H
ARRISON
P. R.: Measures of hope and opti-mism: Assessing positive expectations of the future. In
Measures of per-sonality and social psychological constructs . Elsevier, 2015, pp. 47–73.6[BH19] B
ATTLE
L., H
EER
J.: Characterizing exploratory visual analy-sis: A literature review and evaluation of analytic provenance in tableau.
Computer Graphics Forum 38 , 3 (2019), 145–159. 14[BM93] B
ENYON
D., M
URRAY
D.: Developing adaptive systems to fitindividual aptitudes. In
Proceedings of the 1st international conferenceon Intelligent user interfaces (1993), ACM, pp. 115–121. 1[BMBH16] B
ÖRNER
K., M
ALTESE
A., B
ALLIET
R. N., H
EIMLICH
J.:Investigating aspects of data visualization literacy using 20 informationvisualizations and 273 science museum visitors.
Information Visualiza-tion 15 , 3 (2016), 198–213. 2[BOZ ∗
14] B
ROWN
E. T., O
TTLEY
A., Z
HAO
H., L IN Q., S
OUVENIR
R., E
NDERT
A., C
HANG
R.: Finding waldo: Learning about users fromtheir interactions.
IEEE Transactions on visualization and computergraphics 20 , 12 (2014), 1663–1672. 1, 3, 4, 5, 6, 7, 12, 13, 14[BRBF14] B OY J., R
ENSINK
R. A., B
ERTINI
E., F
EKETE
J.-D.: Aprincipled way of assessing visualization literacy.
IEEE transactions onvisualization and computer graphics 20 , 12 (2014), 1963–1972. 2 [Brö03] B
RÖDER
A.: Decision making with the" adaptive toolbox": in-fluence of environmental structure, intelligence, and working memoryload.
Journal of Experimental Psychology: Learning, Memory, and Cog-nition 29 , 4 (2003), 611. 10[BSC19] B
UCKLEY
J., S
EERY
N., C
ANTY
D.: Investigating the use ofspatial reasoning strategies in geometric problem solving.
InternationalJournal of Technology and Design Education 29 , 2 (2019), 341–362. 8[Car74] C
ARROLL
J. B.:
Psychometric Tests as Cognitive Tasks: ANew’Structure of Intellect’.
Tech. rep., EDUCATIONAL TESTINGSERVICE PRINCETON NJ, 1974. 4, 10[CC97] C
HEN
C., C
ZERWINSKI
M.: Spatial ability and visual naviga-tion: An empirical study.
New Review of Hypermedia and Multimedia 3 ,1 (1997), 67–89. 3, 8, 9, 12[CCH ∗ ARENINI
G., C
ONATI
C., H
OQUE
E., S
TEICHEN
B.,T
OKER
D., E
NNS
J.: Highlighting interventions and user differences:informing adaptive information visualization support. In
Proceedingsof the SIGCHI Conference on Human Factors in Computing Systems (2014), ACM, pp. 1835–1844. 3, 8, 9, 10[CCH ∗ ONATI
C., C
ARENINI
G., H
OQUE
E., S
TEICHEN
B.,T
OKER
D.: Evaluating the impact of user characteristics and differentlayouts on an interactive visualization for decision making.
ComputerGraphics Forum 33 , 3 (2014), 371–380. 3, 8, 9, 10[CGW14] C
OOK
K., G
RINSTEIN
G., W
HITING
M.: The vast challenge:History, scope, and outcomes: An introduction to the special issue, 2014.14[CH07] C
OHEN
C. A., H
EGARTY
M.: Individual differences in use ofexternal visualisations to perform an internal visualisation task.
AppliedCognitive Psychology: The Official Journal of the Society for AppliedResearch in Memory and Cognition 21 , 6 (2007), 701–711. 3, 9, 12[CHD03] C
APON
A., H
ANDLEY
S., D
ENNIS
I.: Working memory andreasoning: An individual differences perspective.
Thinking & Reasoning9 , 3 (2003), 203–244. 10[Che00] C
HEN
C.: Individual differences in a spatial-semantic virtualenvironment.
Journal of the American society for information science51 , 6 (2000), 529–542. 3, 8, 9, 10, 11, 12, 13[CIV ∗
19] C
UR ¸SEU
P. L., I
LIES
R., V
ÎRG ˘A
D., M
ARICU ¸TOIU
L., S
AVA
F. A.: Personality characteristics that are valued in teams: Not al-ways â ˘AIJmore is betterâ ˘A˙I?
International Journal of Psychology 54 ,5 (2019), 638–649. 6[CKT16] C
HARCHAROS
C., K
OKLA
M., T
OMAI
E.: Linking spatialthinking and problem-solving skills of young children.
INSPIRING SCI-ENCE EDUCATION 61 (2016). 8[CLRT17] C
ONATI
C., L
ALLÉ
S., R
AHMAN
M. A., T
OKER
D.: Furtherresults on predicting cognitive abilities for adaptive visualizations. In
IJCAI (2017), pp. 1568–1574. 3, 9, 10, 14[CM08] C
ONATI
C., M
ACLAREN
H.: Exploring the role of individualdifferences in information visualization. In
Proceedings of the workingconference on Advanced visual interfaces (2008), pp. 199–206. 1, 3, 7,8, 9, 10, 12[Cob] C
OBLIS : C
OLOR B LINDNESS S IMULATOR . 14[Cola] C
OLOR O RACLE : https://colororacle.org. Accessed July 15, 2019.URL: https://colororacle.org . 14[Colb] C
OLOR B REWER http://colorbrewer2.org/ . 14[Cor72] C
ORSI
P.: Human memory and the medial temporal lobe regionof the brain (unpublished doctoral dissertation).
McGill University, Mon-treal, Quebec, Canada (1972). 10[Cow08] C
OWAN
N.: What are the differences between long-term, short-term, and working memory?
Progress in brain research 169 (2008),323–338. 10 c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization [CP82] C ACIOPPO
J. T., P
ETTY
R. E.: The need for cognition.
Journalof personality and social psychology 42 , 1 (1982), 116. 4, 7[CP13] C
IORBEA
I., P
ASARICA
F.: The study of the relationship betweenpersonality and academic performance.
Procedia-Social and BehavioralSciences 78 (2013), 400–404. 5[CPFJ96] C
ACIOPPO
J. T., P
ETTY
R. E., F
EINSTEIN
J. A., J
ARVIS
W.B. G.: Dispositional differences in cognitive motivation: The life andtimes of individuals varying in need for cognition.
Psychological bulletin119 , 2 (1996), 197. 7, 8[CPFK84] C
ACIOPPO
J. T., P
ETTY
R. E., F
ENG K AO C.: The efficientassessment of need for cognition.
Journal of personality assessment 48 ,3 (1984), 306–307. 7[CRE ∗
93] C
HARTRAND
J. M., R
OSE
M. L., E
LLIOTT
T. R., M AR - MAROSH
C., C
ALDWELL
S.: Peeling back the onion: Personality, prob-lem solving, and career decision-making style correlates of career inde-cision.
Journal of Career Assessment 1 , 1 (1993), 66–82. 5[CSW55] C
OHEN
A. R., S
TOTLAND
E., W
OLFE
D. M.: An experimen-tal investigation of need for cognition.
The Journal of Abnormal andSocial Psychology 51 , 2 (1955), 291. 7[CWCO19] C
ASHMAN
D., W U Y., C
HANG
R., O
TTLEY
A.: Inferen-tial tasks as a data-rich evaluation method for visualization.
EVIVA-ML:IEEE VIS Workshop on EValuation of Interactive VisuAl Machine Learn-ing systems (2019). 7[CZGR09] C
HANG
R., Z
IEMKIEWICZ
C., G
REEN
T. M., R
IBARSKY
W.: Defining insight for visual analytics.
IEEE Computer Graphics andApplications 29 , 2 (2009), 14–17. 13[DC80] D
ANEMAN
M., C
ARPENTER
P. A.: Individual differences inworking memory and reading.
Journal of Memory and Language 19 , 4(1980), 450. 10[DC17] D
ABEK
F., C
ABAN
J. J.: A grammar-based approach for mod-eling user interactions and generating suggestions during the data explo-ration process.
IEEE transactions on visualization and computer graph-ics 23 , 1 (2017), 41–50. 14[DGP12] D E Y OUNG
C. G., G
RAZIOPLENE
R. G., P
ETERSON
J. B.:From madness to genius: The openness/intellect trait domain as a para-doxical simplex.
Journal of Research in Personality 46 , 1 (2012), 63–78.5[DGSO06] D
RISKELL
J. E., G
OODWIN
G. F., S
ALAS
E., O’S
HEA
P. G.: What makes a good team player? personality and team effec-tiveness.
Group Dynamics: Theory, Research, and Practice 10 , 4 (2006),249. 6[Dia13] D
IAMOND
A.: Executive functions.
Annual review of psychology64 (2013), 135–168. 10[Dis03] D
ISETH
Å.: Personality and approaches to learning as predictorsof academic achievement.
European Journal of personality 17 , 2 (2003),143–155. 6[DJS ∗
09] D OU W., J
EONG
D. H., S
TUKES
F., R
IBARSKY
W., L
IPFORD
H. R., C
HANG
R.: Recovering reasoning processes from user interac-tions.
IEEE Computer Graphics and Applications 29 , 3 (2009), 52–61.2[DM96] D
ANEMAN
M., M
ERIKLE
P. M.: Working memory and lan-guage comprehension: A meta-analysis.
Psychonomic bulletin & review3 , 4 (1996), 422–433. 10[DMBM09] D
ORNBURG
C. C., M
ATZEN
L. E., B
AUER
T. L., M C N A - MARA
L. A.: Working memory load as a novel tool for evaluating visualanalytics. In (2009), IEEE, pp. 217–218. 8[DMOGP11] D Â ˘A ´ZZ URILLA
T. J., M
AYDEU -O LIVARES
A.,G
ALLARDO -P UJOL
D.: Predicting social problem solving usingpersonality traits.
Personality and individual Differences 50 , 2 (2011),142–147. 6[DOBL06] D
ONNELLAN
M. B., O
SWALD
F. L., B
AIRD
B. M., L
UCAS
R. E.: The mini-ipip scales: tiny-yet-effective measures of the big fivefactors of personality.
Psychological assessment 18 , 2 (2006), 192. 5 [DQP07] D E Y OUNG
C. G., Q
UILTY
L. C., P
ETERSON
J. B.: Betweenfacets and domains: 10 aspects of the big five.
Journal of personality andsocial psychology 93 , 5 (2007), 880. 5[DW96] D
ILLON
A., W
ATSON
C.: User analysis in hci: the historicallesson from individual differences research. Elsevier. 1[EDH76] E
KSTROM
R. B., D
ERMEN
D., H
ARMAN
H. H.:
Manual forkit of factor-referenced cognitive tests , vol. 102. Educational testing ser-vice Princeton, NJ, 1976. 4, 8, 9, 10, 11[EG85] E
GAN
D. E., G
OMEZ
L. M.: Assaying, isolating, and accom-modating individual differences in learning a complex skill.
Individualdifferences in cognition 2 (1985), 174–217. 15[Far66] F
ARLEY
F.: Individual differences in solution time in error-freeproblem solving.
British Journal of Social and Clinical Psychology 5 , 4(1966), 306–309. 5[FC83] F
INDLEY
M. J., C
OOPER
H. M.: Locus of control and academicachievement: A literature review.
Journal of personality and social psy-chology 44 , 2 (1983), 419. 6[FDF ∗
86] F
LETCHER
G. J., D
ANILOVICS
P., F
ERNANDEZ
G., P
ETER - SON
D., R
EEDER
G. D.: Attributional complexity: An individual dif-ferences measure.
Journal of Personality and Social Psychology 51 , 4(1986), 875. 8[FJ96] F
ISHER
K., J
OHNSTION
M.: Experimental manipulation of per-ceived control and its effect on disability.
Psychology and Health 11 , 5(1996), 657–669. 7[FTES13] F
ROESE
M.-E., T
ORY
M., E
VANS
G.-W., S
HRIKHANDE
K.:Evaluation of static and dynamic visualization training approaches forusers with different spatial abilities.
IEEE transactions on visualizationand computer graphics 19 , 12 (2013), 2810–2817. 3, 8, 9, 12[FV09] F
UKUDA
K., V
OGEL
E. K.: Human variation in overriding at-tentional capture.
Journal of Neuroscience 29 , 27 (2009), 8726–8733.10[GBPM06] G
IFFORD
D. D., B
RICENO -P ERRIOTT
J., M
IANZO
F.: Lo-cus of control: Academic achievement and retention in a sample of uni-versity first-year students.
Journal of College Admission 191 (2006),18–25. 6[GC17] G
AJOS
K. Z., C
HAUNCEY
K.: The influence of personality traitsand cognitive load on the use of adaptive user interfaces. In
Proceed-ings of the 22nd International Conference on Intelligent User Interfaces (2017), ACM, pp. 301–306. 1[Gen14] G
ENSOWSKI
M.: Personality, iq, and lifetime earnings. IZADiscussion Paper. 5[GF10] G
REEN
T. M., F
ISHER
B.: Towards the personal equation of in-teraction: The impact of personality factors on visual analytics interfaceinteraction. In (2010), IEEE, pp. 203–210. 1, 3, 4, 5, 6, 12, 13[GF12] G
REEN
T. M., F
ISHER
B.: Impact of personality factors on in-terface interaction and the development of user profiles: Next steps in thepersonal equation of interaction.
Information Visualization 11 , 3 (2012),205–221. 1, 12[GJF10] G
REEN
T. M., J
EONG
D. H., F
ISHER
B.: Using personalityfactors to predict interface learning performance. In (2010), IEEE, pp. 1–10. 1,3, 4, 5, 6, 11, 13[Gol93] G
OLDBERG
L. R.: The structure of phenotypic personality traits.
American psychologist 48 , 1 (1993), 26. 4, 5[GSC16] G
OTZ
D., S UN S., C AO N.: Adaptive contextualization: Com-bating bias during high-dimensional visualization and data selection. In
Proceedings of the 21st International Conference on Intelligent User In-terfaces (2016), pp. 85–95. 15[GSL04] G
IGNAC
G. E., S
TOUGH
C., L
OUKOMITIS
S.: Openness, intel-ligence, and self-report intelligence.
Intelligence 32 , 2 (2004), 133–143.5 c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization [GW09] G OTZ
D., W EN Z.: Behavior-driven visualization recommenda-tion. In
Proceedings of the 14th international conference on Intelligentuser interfaces (2009), ACM, pp. 315–324. 15[GZ09] G
OTZ
D., Z
HOU
M. X.: Characterizing users’ visual analyticactivity for insight provenance.
Information Visualization 8 , 1 (2009),42–55. 14[HB10] H
EER
J., B
OSTOCK
M.: Crowdsourcing graphical perception:using mechanical turk to assess visualization design. In
Proceedings ofthe SIGCHI conference on human factors in computing systems (2010),pp. 203–212. 13[Hei10] H
EINSTROM
J.:
From fear to flow: personality and informationinteraction . Elsevier, 2010. 6[HHL11] H
AKIMI
S., H
EJAZI
E., L
AVASANI
M. G.: The relation-ships between personality traits and studentsâ ˘A ´Z academic achievement.
Procedia-Social and Behavioral Sciences 29 (2011), 836–845. 5, 6[HJW02] H
INSON
J. M., J
AMESON
T. L., W
HITNEY
P.: Somatic mark-ers, working memory, and decision making.
Cognitive, Affective, & Be-havioral Neuroscience 2 , 4 (2002), 341–353. 10[HJW03] H
INSON
J. M., J
AMESON
T. L., W
HITNEY
P.: Impulsive deci-sion making and working memory.
Journal of Experimental Psychology:Learning, Memory, and Cognition 29 , 2 (2003), 298. 10[Hor99] H
ORVITZ
E.: Principles of mixed-initiative user interfaces. In
Proceedings of the SIGCHI conference on Human Factors in ComputingSystems (1999), ACM, pp. 159–166. 15[HTA ∗
14] H
UANG
D., T
ORY
M., A
SENIERO
B. A., B
ARTRAM
L.,B
ATEMAN
S., C
ARPENDALE
S., T
ANG
A., W
OODBURY
R.: Personalvisualization and personal visual analytics.
IEEE Transactions on Visu-alization and Computer Graphics 21 , 3 (2014), 420–433. 14[Hul28] H
ULL
C. L.: Aptitude testing. World Book Company. 1[HVRT12] H
AZRATI -V IARI
A., R AD A. T., T
ORABI
S. S.: The effect ofpersonality traits on academic performance: The mediating role of aca-demic motivation.
Procedia-Social and Behavioral Sciences 32 (2012),367–371. 5[IB15] I
SPAS
D., B
ORMAN
W. C.: Personnel selection, psychology of.Elsevier. 8[IMB04] I
OANNOU
M. C., M
OGG
K., B
RADLEY
B. P.: Vigilance forthreat: Effects of anxiety and defensiveness.
Personality and IndividualDifferences 36 , 8 (2004), 1879–1891. 5[Jen15] J
ENSEN
M.: Personality traits, learning and academic achieve-ments.
Journal of Education and Learning 4 , 4 (2015), 91–118. 5[JGPC92] J
OHNSTON
M., G
ILBERT
P., P
ARTRIDGE
C., C
OLLINS
J.:Changing perceived control in patients with physical disabilities: An in-tervention study with patients receiving rehabilitation.
British Journal ofClinical Psychology 31 , 1 (1992), 89–94. 7[Joh14] J
OHNSON
J. A.: Measuring thirty facets of the five factor modelwith a 120-item public domain inventory: Development of the ipip-neo-120.
Journal of Research in Personality 51 (2014), 78–89. 5[JRSO14] J
ERONIMUS
B. F., R
IESE
H., S
ANDERMAN
R., O
RMEL
J.:Mutual reinforcement between neuroticism and life experiences: A five-wave, 16-year study to test reciprocal causation.
Journal of personalityand social psychology 107 , 4 (2014), 751. 5[Kim00] K
IMURA
D.:
Sex and cognition . MIT press, 2000. 8[Kla97] K
LAUER
K. C.: Working memory involvement in propositionaland spatial reasoning.
Thinking & Reasoning 3 , 1 (1997), 9–47. 10[KLWG13] K
ARWOWSKI
M., L
EBUDA
I., W
ISNIEWSKA
E.,G
RALEWSKI
J.: Big five personality traits as the predictors ofcreative self-efficacy and creative personal identity: Does gender matter?
The Journal of Creative Behavior 47 , 3 (2013), 215–232. 6[Kob04] K
OBSA
A.: User experiments with tree visualization systems. In
IEEE Symposium on Information Visualization (2004), IEEE, pp. 9–16.1 [KZ10] K
OSARA
R., Z
IEMKIEWICZ
C.: Do mechanical turks dream ofsquare pie charts? In
Proceedings of the 3rd BELIV’10 Workshop: Be-yond time and errors: Novel evaluation methods for information visual-ization (2010), pp. 63–70. 13[LC19] L
ALLÉ
S., C
ONATI
C.: The role of user differences in customiza-tion: a case study in personalization for infovis-based content. In
Pro-ceedings of the 24th International Conference on Intelligent User Inter-faces (2019), pp. 329–339. 3, 10, 13[LCC17] L
ALLÉ
S., C
ONATI
C., C
ARENINI
G.: Impact of individualdifferences on user experience with a real-world visualization interfacefor public engagement. In
Proceedings of the 25th Conference on UserModeling, Adaptation and Personalization (2017), ACM, pp. 369–370.9, 10, 13[LDH ∗
19] L EE D. J.-L., D EV H., H U H., E
LMELEEGY
H.,P
ARAMESWARAN
A.: Avoiding drill-down fallacies with vispilot: As-sisted exploration of data subsets. In
Proceedings of the 24th Interna-tional Conference on Intelligent User Interfaces (2019), pp. 186–196.15[LH14] L IU Z., H
EER
J.: The effects of interactive latency on exploratoryvisual analysis.
IEEE transactions on visualization and computer graph-ics 20 , 12 (2014), 2122–2131. 14[LKK16] L EE S., K IM S.-H., K
WON
B. C.: Vlat: Development of avisualization literacy assessment test.
IEEE transactions on visualizationand computer graphics 23 , 1 (2016), 551–560. 2[LOR ∗
11] L E H., O H I.-S., R
OBBINS
S. B., I
LIES
R., H
OLLAND
E.,W
ESTRICK
P.: Too much of a good thing: Curvilinear relationships be-tween personality traits and job performance.
Journal of Applied Psy-chology 96 , 1 (2011), 113. 6[LSLG03] L
OUNSBURY
J. W., S
UNDSTROM
E., L
OVELAND
J. L., G IB - SON
L. W.: Broad versus narrow personality traits in predicting aca-demic performance of adolescents.
Learning and individual differences14 , 1 (2003), 65–75. 6[LTC19] L
ALLÉ
S., T
OKER
D., C
ONATI
C.: Gaze-driven adaptive in-terventions for magazine-style narrative visualizations. arXiv preprintarXiv:1909.01379 (2019). 1[LTCC15] L
ALLÉ
S., T
OKER
D., C
ONATI
C., C
ARENINI
G.: Predic-tion of users’ learning curves for adaptation while using an informationvisualization. In
Proceedings of the 20th International Conference onIntelligent User Interfaces (2015), pp. 357–368. 3, 9, 10, 11, 13[LV97] L
UCK
S. J., V
OGEL
E. K.: The capacity of visual working mem-ory for features and conjunctions.
Nature 390 , 6657 (1997), 279–281. 10[LV13] L
UCK
S. J., V
OGEL
E. K.: Visual working memory capacity:from psychophysics and neurobiology to individual differences.
Trendsin cognitive sciences 17 , 8 (2013), 391–400. 10[MA03] M
OHAMMED
S., A
NGELL
L. C.: Personality heterogeneityin teams: Which differences make a difference for team performance?
Small group research 34 , 6 (2003), 651–677. 6[MD16] M
ARVEL
C. L., D
ESMOND
J. E.: The cerebellum and verbalworking memory. In
The linguistic cerebellum . Elsevier, 2016, pp. 51–62. 10[MDF12] M
ICALLEF
L., D
RAGICEVIC
P., F
EKETE
J.-D.: Assessing theeffect of visualizations on bayesian reasoning through crowdsourcing.
IEEE transactions on visualization and computer graphics 18 , 12 (2012),2536–2545. 3, 11, 12, 13[Mel82] M
ELTZER
L. J.: Visual perception: Stage one of a long-terminvestigation into cognitive components of achievement.
British Journalof Educational Psychology 52 , 2 (1982), 144–154. 9[MGP60] M
ILLER
G. A., G
ALANTER
E., P
RIBRAM
K. H.: Plans andthe structure of behavior. Henry Holt and Co. 10[MHCV19] M
ILLECAMP
M., H
TUN
N. N., C
ONATI
C., V
ERBERT
K.:To explain or not to explain: the effects of personal characteristics whenexplaining music recommendations. In
IUI (2019), pp. 397–407. 3, 7, 8,12, 13 c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization [MMWL14] M ALL
J. T., M
OREY
C. C., W
OLFF
M. J., L
EHNERT
F.:Visual selective attention is equally functional for individuals with lowand high working memory capacity: Evidence from accuracy and eyemovements.
Attention, Perception, & Psychophysics 76 , 7 (2014), 1998–2014. 10[MR93] M
AIN
J. D., R
OWE
M. B.: The relation of locus-of-controlorientation and task structure to problem-solving performance of sixth-grade student pairs.
Journal of Research in Science Teaching 30 , 4(1993), 401–426. 6[MS99] M
IYAKE
A., S
HAH
P.:
Models of working memory: Mecha-nisms of active maintenance and executive control . Cambridge Univer-sity Press, 1999. 10[MSDL15] M
YSZKOWSKI
N., S
TORME
M., D
AVILA
A., L
UBART
T.:Managerial creative problem solving and the big five personality traits:Distinguishing divergent and convergent abilities.
Journal of manage-ment development 34 , 6 (2015), 674–684. 5[Mun09] M
UNZNER
T.: A nested process model for visualization designand validation.
IEEE Visualization and Computer Graphics 15 , 6 (2009),921–928. 14[NALB13] N OV O., A
RAZY
O., L
ÓPEZ
C., B
RUSILOVSKY
P.: Explor-ing personality-targeted ui design in online social participation systems.In
Proceedings of the SIGCHI Conference on Human Factors in Com-puting Systems (2013), ACM, pp. 361–370. 1[ND86] N
ORMAN
D. A., D
RAPER
S. W.:
User centered system design:New perspectives on human-computer interaction . CRC Press, 1986. 15[NZ15] N
IESS
C., Z
ACHER
H.: Openness to experience as a predictorand outcome of upward job changes into managerial and professionalpositions.
PloS one 10 , 6 (2015), e0131115. 5[O‘C02] O‘C
ONNOR
B. P.: A quantitative review of the comprehensive-ness of the five-factor model in relation to popular personality invento-ries.
Assessment 9 , 2 (2002), 188–203. 4[OCZC12] O
TTLEY
A., C
ROUSER
R. J., Z
IEMKIEWICZ
C., C
HANG
R.:Priming locus of control to affect performance. In (2012), IEEE,pp. 237–238. 4, 11, 12[OCZC15] O
TTLEY
A., C
ROUSER
R. J., Z
IEMKIEWICZ
C., C
HANG
R.: Manipulating and controlling for personality effects on visualiza-tion tasks.
Information Visualization 14 , 3 (2015), 223–233. 1, 3, 7, 12,13[OGW19] O
TTLEY
A., G
ARNETT
R., W AN R.: Follow The Clicks :Learning and Anticipating Mouse Interactions During Exploratory DataAnalysis. In
Computer Graphics Forum (2019), Wiley Online Library.14[ONDM17] O
RJI
R., N
ACKE
L. E., D I M ARCO
C.: Towardspersonality-driven persuasive health games and gamified systems. In
Proceedings of the 2017 CHI Conference on Human Factors in Comput-ing Systems (2017), ACM, pp. 1015–1027. 1[OPH ∗
15] O
TTLEY
A., P
ECK
E. M., H
ARRISON
L. T., A
FERGAN
D.,Z
IEMKIEWICZ
C., T
AYLOR
H. A., H AN P. K., C
HANG
R.: Improvingbayesian reasoning: The effects of phrasing, visualization, and spatialability.
IEEE transactions on visualization and computer graphics 22 , 1(2015), 529–538. 1, 3, 8, 9, 11, 12, 13[OS15] O ˘GUZ
A., S
ARIÇAM
H.: The relationship between critical think-ing disposition and locus of control in pre-service teachers.
Journal ofEducation and Training Studies 4 , 2 (2015), 182–192. 6[OYC15] O
TTLEY
A., Y
ANG
H., C
HANG
R.: Personality as a predictorof user strategy: How locus of control affects search strategies on treevisualizations. In
Proceedings of the 33rd Annual ACM Conference onHuman Factors in Computing Systems (2015), ACM, pp. 3251–3254. 1,3, 4, 7, 11, 12, 13, 14[PCI10] P
AOLACCI
G., C
HANDLER
J., I
PEIROTIS
P. G.: Running ex-periments on amazon mechanical turk.
Judgment and Decision making5 , 5 (2010), 411–419. 13 [Poc91] P
OCIUS
K. E.: Personality factors in human-computer interac-tion: A review of the literature.
Computers in Human Behavior 7 , 3(1991), 103–135. 1[PVTRR06] P
EETERS
M. A., V AN T UIJL
H. F., R
UTTE
C. G., R EY - MEN
I. M.: Personality and team performance: a meta-analysis.
Euro-pean Journal of Personality: Published for the European Association ofPersonality Psychology 20 , 5 (2006), 377–396. 6[PYH ∗
12] P
ECK
E. M., Y
UKSEL
B. F., H
ARRISON
L., O
TTLEY
A.,C
HANG
R.: Towards a 3-dimensional model of individual cognitivedifferences: position paper. In
Proceedings of the 2012 BELIV Work-shop: Beyond Time and Errors-Novel Evaluation Methods for Visualiza-tion (2012), pp. 1–6. 2[RC03] R
OTHMANN
S., C
OETZER
E. P.: The big five personality di-mensions and job performance.
SA Journal of Industrial Psychology 29 ,1 (2003), 68–74. 6[RJF ∗
09] R
OBERTS
B., J
ACKSON
J., F
AYARD
J., E
DMONDS
G.,M
EINTS
J.: Conscientiousness in mlr hoyle (ed.), handbook of individ-ual differences in social behavior (pp. 369-381).
New York, NY: Guilford (2009). 6[Rot66] R
OTTER
J. B.: Generalized expectancies for internal versus ex-ternal control of reinforcement.
Psychological monographs: Generaland applied 80 , 1 (1966), 1. 4, 6[Rot75] R
OTTER
J. B.: Some problems and misconceptions related to theconstruct of internal versus external control of reinforcement.
Journal ofconsulting and clinical psychology 43 , 1 (1975), 56. 4, 6[Rot90] R
OTTER
J. B.: Internal versus external control of reinforcement:A case history of a variable.
American psychologist 45 , 4 (1990), 489. 4,6[RS13] R AY -S UBRAMANIAN
C.: Visual-spatial ability.
Encyclopedia ofAutism Spectrum Disorders (2013), 3326–3328. 4, 8[SBF04] S
WANSON
H. L., B
EEBE -F RANKENBERGER
M.: The rela-tionship between working memory and mathematical problem solvingin children at risk and not at risk for serious math difficulties.
Journal ofEducational Psychology 96 , 3 (2004), 471. 10[SCC13] S
TEICHEN
B., C
ARENINI
G., C
ONATI
C.: User-adaptive infor-mation visualization: using eye gaze data to infer visualization tasks anduser cognitive abilities. In
Proceedings of the 2013 international confer-ence on Intelligent user interfaces (2013), ACM, pp. 317–328. 3, 8, 9,10, 11, 12, 13, 14[SCC14] S
TEICHEN
B., C
ONATI
C., C
ARENINI
G.: Inferring visual-ization task properties, user performance, and user cognitive abilitiesfrom eye gaze data.
ACM Transactions on Interactive Intelligent Sys-tems (TiiS) 4 , 2 (2014), 1–29. 3, 8, 10, 13[SCJME09] S
UTIN
A. R., C
OSTA J R P. T., M
IECH
R., E
ATON
W. W.:Personality and career success: Concurrent and longitudinal relations.
European Journal of Personality: Published for the European Associa-tion of Personality Psychology 23 , 2 (2009), 71–84. 6[SDL ∗
18] S
ORBY
S. A., D
UFFY
G., L
ONEY
N., P
EREZ
L.,
ET AL .:Spatial skills and their correlation with engineering problem-solving.In (2018), Engineers Australia, p. 10. 8[Shn96] S
HNEIDERMAN
B.: The eyes have it: A task by data type taxon-omy for information visualizations. In
Proceedings 1996 IEEE sympo-sium on visual languages (1996), IEEE, pp. 336–343. 14[Shu91] S
HUTE
V. J.: Who is likely to acquire programming skills?
Jour-nal of educational Computing research 7 , 1 (1991), 1–24. 9[SLC ∗
20] S
HEIDIN
J., L
ANIR
J., C
ONATI
C., T
OKER
D., K
UFLIK
T.:The effect of user characteristics in time series visualizations. In
Pro-ceedings of the 25th International Conference on Intelligent User Inter-faces (2020), pp. 380–389. 3, 12[Spe63] S
PERLING
G.: A model for visual memory tasks.
Human factors5 , 1 (1963), 19–31. 4, 9 c (cid:13) (cid:13) . Liu, R. J. Crouser & A. Ottley / Survey on Individual Differences in Visualization [TC05] T HOMAS
J. J., C
OOK
K. A.:
Illuminating the path: The researchand development agenda for visual analytics . IEEE Computer SocietyPress, 2005. 15[TCC19] T
OKER
D., C
ONATI
C., C
ARENINI
G.: Gaze analysis of usercharacteristics in magazine style narrative visualizations.
User Modelingand User-Adapted Interaction 29 , 5 (2019), 977–1011. 3, 7, 8, 9, 10, 11,12, 13, 14[TCCH12] T
OKER
D., C
ONATI
C., C
ARENINI
G., H
ARATY
M.: To-wards adaptive information visualization: on the influence of user char-acteristics. In
International conference on user modeling, adaptation,and personalization (2012), Springer, pp. 274–285. 3, 8, 9, 10, 11, 13[TCSC13] T
OKER
D., C
ONATI
C., S
TEICHEN
B., C
ARENINI
G.: Indi-vidual user characteristics and information visualization: connecting thedots through eye tracking. In proceedings of the SIGCHI Conference onHuman Factors in Computing Systems (2013), ACM, pp. 295–304. 3, 8,9, 10, 11, 12, 13[TE89] T
URNER
M. L., E
NGLE
R. W.: Is working memory capacity taskdependent?
Journal of memory and language 28 , 2 (1989), 127–154. 10[Tou12] T
OUGH
P.:
How children succeed: Grit, curiosity, and the hiddenpower of character . Houghton Mifflin Harcourt, 2012. 6[Uzi06] U
ZIEL
L.: The extraverted and the neurotic glasses are of differ-ent colors.
Personality and Individual Differences 41 , 4 (2006), 745–754.5[vDM16]
VAN D UN K., M
ARIËN
P.: Cerebellar-induced aphasia andrelated language disorders. In
The Linguistic Cerebellum . Elsevier, 2016,pp. 107–133. 10[VH15] V
ANDER P LAS
S., H
OFMANN
H.: Spatial reasoning and datadisplays.
IEEE Transactions on Visualization and Computer Graphics22 , 1 (2015), 459–468. 3, 8, 9, 12[VHW87] V
ICENTE
K. J., H
AYES
B. C., W
ILLIGES
R. C.: Assayingand isolating individual differences in searching a hierarchical file sys-tem.
Human factors 29 , 3 (1987), 349–359. 1, 3, 4, 8, 9, 12, 13, 15[VK78] V
ANDENBERG
S. G., K
USE
A. R.: Mental rotations, a group testof three-dimensional spatial visualization.
Perceptual and motor skills47 , 2 (1978), 599–604. 8[Voo97] V
OOGHT
A. V. G. D.: Working memory constraints on linearreasoning with spatial and temporal contents.
The Quarterly Journal ofExperimental Psychology: Section A 50 , 4 (1997), 803–820. 10[VST05] V
ELEZ
M. C., S
ILVER
D., T
REMAINE
M.: Understanding vi-sualization through spatial ability differences. In
VIS 05. IEEE Visual-ization, 2005. (2005), IEEE, pp. 511–518. 3, 8, 9, 12[VWL01] V
OGEL
E. K., W
OODMAN
G. F., L
UCK
S. J.: Storage offeatures, conjunctions, and objects in visual working memory.
Journalof Experimental Psychology: Human Perception and Performance 27 , 1(2001), 92. 10[WB00] W
OOLHOUSE
L. S., B
AYNE
R.: Personality and the use of in-tuition: Individual differences in strategy and performance on an implicitlearning task.
European Journal of Personality 14 , 2 (2000), 157–169. 1[WBFE17] W
ALL
E., B
LAHA
L. M., F
RANKLIN
L., E
NDERT
A.: Warn-ing, bias may occur: A proposed approach to detecting cognitive bias ininteractive visual analytics. In (2017), IEEE, pp. 104–115. 14,15[WHA ∗
02] W
ANZEL
K. R., H
AMSTRA
S. J., A
NASTAKIS
D. J., M AT - SUMOTO
E. D., C
USIMANO
M. D.: Effect of visual-spatial abilityon learning of spatially-complex surgical skills.
The lancet 359 , 9302(2002), 230–231. 8[Wil04] W
ILLIAMS
M.:
Questvis and mdsteer: The visualization of high-dimensional environmental sustainability data . PhD thesis, University ofBritish Columbia, 2004. 9[WL17] W
ANG
L., L V M.: Internal-external locus of control scale.
En-cyclopedia of Personality and Individual Differences (2017), 1–4. 4 [WLB09] W AI J., L
UBINSKI
D., B
ENBOW
C. P.: Spatial ability for stemdomains: Aligning over 50 years of cumulative psychological knowl-edge solidifies its importance.
Journal of Educational Psychology 101 ,4 (2009), 817. 8[WR17] W
ILT
J., R
EVELLE
W.: Extraversion. 5[YaKSJ07] Y I J. S., AH K ANG
Y., S
TASKO
J. T., J
ACKO
J. A.: Towarda deeper understanding of the role of interaction in information visu-alization.
IEEE Transactions on Visualization and Computer Graphics(TVCG) 13 , 6 (2007), 1224–1231. 14[Yi12] Y I J. S.: Implications of individual differences on evaluating in-formation visualization techniques.
International Journal of Human-Computer Studies 45 , 6 (2012), 619–637. 1[YLM18] Y
OUNG
C. J., L
EVINE
S. C., M IX K. S.: The connectionbetween spatial and mathematical ability across development.
Frontiersin psychology 9 (2018). 8[ZCY ∗
11] Z
IEMKIEWICZ
C., C
ROUSER
R. J., Y
AUILLA
A. R., S U S. L., R
IBARSKY
W., C
HANG
R.: How locus of control influencescompatibility with visualization style. In (2011), IEEE, pp. 81–90.1, 2, 4, 5, 6, 7, 12[ZF98] Z
HOU
M. X., F
EINER
S. K.: Visual task characterization for au-tomated visual discourse synthesis. In
Proceedings of the SIGCHI Con-ference on Human Factors in Computing Systems (CHI) (1998), ACMPress/Addison-Wesley Publishing Co., pp. 392–399. 14[ZK09] Z
IEMKIEWICZ
C., K
OSARA
R.: Preconceptions and individualdifferences in understanding visual metaphors. In
Computer GraphicsForum (2009), vol. 28, Wiley Online Library, pp. 911–918. 3, 5, 6, 8, 9,11, 12[ZOC ∗ IEMKIEWICZ
C., O
TTLEY
A., C
ROUSER
R. J.,C
HAUNCEY
K., S U S. L., C
HANG
R.: Understanding visualiza-tion by understanding individual users.
IEEE computer graphics andapplications 32 , 6 (2012), 88–94. 1[ZOC ∗ IEMKIEWICZ
C., O
TTLEY
A., C
ROUSER
R. J., Y
AUILLA
A. R., S U S. L., R
IBARSKY
W., C
HANG
R.: How visualization layoutrelates to locus of control and other personality factors.
IEEE transac-tions on visualization and computer graphics 19 , 7 (2012), 1109–1121.1, 2, 3, 4, 5, 6, 11, 12, 13 c (cid:13) (cid:13)(cid:13)