Exploring How Personality Models Information Visualization Preferences
Tomás Alves, Bárbara Ramalho, Joana Henriques-Calado, Daniel Gonçalves, Sandra Gama
EExploring How Personality Models Information Visualization Preferences
Toms Alves * , Brbara Ramalho † , Daniel Gonalves ‡ , and Sandra Gama § INESC-ID and Instituto Superior Tcnico,University of Lisbon, Lisbon, Portugal
Joana Henriques-Calado ¶ CICPSI, Faculdade de Psicologia,Universidade de Lisboa, Lisboa, PortugalFigure 1: The three different information visualization contexts that we address in our study. (A) the hierarchy context which isstudied with the treemap, circular packing, sunburst, and Sankey diagram idioms. (B) evolution over time context, where we applyline charts with and without points, and an area chart. Finally, (C) comparison context, which includes the radar chart, word cloud,horizontal and vertical bar charts, and pie chart idioms. A BSTRACT
Recent research on information visualization has shown how in-dividual differences act as a mediator on how users interact withvisualization systems. We focus our exploratory study on whetherpersonality has an effect on user preferences regarding idioms usedfor hierarchy, evolution over time, and comparison contexts. Specif-ically, we leverage all personality variables from the Five-FactorModel and the three dimensions from Locus of Control (LoC) withcorrelation and clustering approaches. The correlation-based methodsuggested that Neuroticism, Openness to Experience, Agreeable-ness, several facets from each trait, and the External dimensionsfrom LoC mediate how much individuals prefer certain idioms. Inaddition, our results from the cluster-based analysis showed thatNeuroticism, Extraversion, Conscientiousness, and all dimensionsfrom LoC have an effect on preferences for idioms in hierarchyand evolution contexts. Our results support the incorporation ofin-depth personality synergies with InfoVis into the design pipelineof visualization systems.
Index Terms:
Human-centered computing—Human computer in-teraction (HCI)—HCI design and evaluation methods—User studies;Human-centered computing—Visualization—Visualization designand evaluation methods
NTRODUCTION
Individual differences have shown promise as an adaptation met-ric of information visualization systems to tackle the limitationsof one-size-fits-all approaches [5, 15, 24]. The inclusion of these * e-mail: [email protected] † e-mail: [email protected] ‡ e-mail: [email protected] § e-mail: [email protected] ¶ e-mail: [email protected] factors empowers developers with guidelines on how individualcharacteristics impact human-computer interaction. Among severalpsychological constructs that differentiate individuals such as cog-nitive bias or abilities, personality stands as an established strongmediator given its stability throughout adulthood [21]. Comparedto other well studied individual cognitive traits (e.g. spatial abilityand visual working memory), promising results regarding the rela-tionship between personality traits and information visualization arefew [20]. In order to bridge this gap, we focus on two of the mostextensively studied personality models in our research field: theFive-Factor Model (FFM) [6] and the Locus of Control (LoC) [16].Although performance metrics such as speed or accuracy areimportant to address while users perform tasks in information visu-alization systems (e.g. [25, 26]), we believe that there is a lack offindings regarding how personality affects user preferences for infor-mation visualization techniques. Weighting how personality has aneffect on user preferences, our study focuses on whether personalitypromotes user preferences regarding idioms used for hierarchy,evolution over time, and comparison contexts , as these contextshave been frequently applied in state-of-the-art research [4, 20, 31].In particular, we address all traits and their facets from the FFM,and the three dimensions from LoC to provide an in-depth analy-sis. We take two distinct approaches to study this relationship: (i)correlation-based analysis – where we investigate whether a person-ality variable had correlations with the user preference regarding anidiom – and (ii) cluster-based analysis – where we aggregate indi-viduals by common characteristics and extract preference patternsfrom each group. Our preliminary results suggest that personalityhas an effect on user preferences with both types of analysis. ELATED W ORK
The Five-Factor Model consists of five general dimensions to de-scribe personality and 30 subdimensions (facets) (Table 1): (i)Neuroticism distinguishes “the stability of emotions and even-temperedness from negative emotionality” [11]; (ii) Extraver-sion suggests “a lively approach toward the social and materialworld” [11]; (iii) Openness to Experience describes “the whole- a r X i v : . [ c s . H C ] S e p able 1: Traits and their facets of the Five-Factor Model. Trait Facets
Neuroticism (N) Anxiety (N1), Anger (N2), Depression (N3), Self-consciousness (N4), Immoderation (N5), Vulnerability (N6)Extraversion (E) Friendliness (E1), Gregariousness (E2), Assertiveness (E3), Activity level (E4), Excitement-seeking (E5), Cheerfulness (E6)Openness to Experience (O) Imagination (O1), Artistic interests (O2), Emotionality (O3), Adventurousness (O4), Intellect (O5), Liberalism (O6)Agreeableness (A) Trust (A1), Morality (A2), Altruism (A3), Cooperation (A4), Modesty (A5), Sympathy (A6)Conscientiousness (C) Self-efficacy (C1), Orderliness (C2), Dutifulness (C3), Achievement-striving (C4), Self-discipline (C5), Cautiousness (C6) ness and complexity of an individual’s psychological and experi-ential life” [11]; (iv) Agreeableness distinguishes “pro-social andcommunal orientation toward others from antagonism” [11]; and(v) Conscientiousness suggests “self-use of socially prescribed re-straints that facilitate goal completion, following norms and rules,and prioritizing tasks” [11]. Ziemkiewicz and Kosara [35] foundthat high Openness to Experience led individuals to be faster whilesolving problems related to hierarchical visualizations that includeconflicting visual and verbal metaphors. Furthermore, Ziemkiewiczet al. [36] concluded that neurotic individuals attained high accuracyon hierarchical search tasks. Introverted participants were moreaccurate in answering the questions posed by the tasks. Other con-tributions [2, 3, 10, 24, 34] have addressed the traits of Neuroticismand Extraversion. Results have shown how these traits have aneffect on task performance metrics such as the time to completea task [10, 24] and accuracy [24]. Additionally, Neuroticism andOpenness to Experience exhibited an effect on the attractivenessand dependability ratings from participants regarding driver statevisualization systems [2].The Locus of Control orientations are described as two differentaspects, which are distinguished by different reinforcements. Whileinternal LoC is related with internal reinforcement because the valueof an individual is heightened by some event or environment, exter-nal LoC is linked to external reinforcement since it addresses howsome event or environment yields benefit for the group or culture towhich the individual belongs to [14]. Furthermore, external LoC canbe differentiated in two types:
Powerful Others – believe in an or-dered world controlled by powerful others – and
Chance – considerthe world as unordered and chaotic [18]. Several studies have shownhow LoC is related to search performance across hierarchical [10],time series [31], and item comparison [4] visualization designs, vi-sualization use [34, 36], and behavioural patterns [26]. Although
Internals are significantly faster than
Externals when performingprocedural tasks (search tasks to locate items) [10],
Externals arefaster and more accurate than
Internals regarding inferential taskssuch as comparing two items [34, 36]. In addition,
Internals areusually faster than
Externals in image-based search tasks [3].Although performance metrics such as speed or accuracy areimportant to address, there are strong results regarding user prefer-ences in information visualization. Ziemkiewicz et al. [36] focus onNeuroticism, Extraversion, and the LoC, while Lall and Conati [15]address the latter. In contrast, Toker et al. [32] did not addresspersonality. Nevertheless, research has not found effects for Agree-ableness or Conscientiousness [20], and FFM traits’ facets havebeen neglected. In our study, we propose an extension of the state-of-the-art research by including the remaining FFM traits and theirfacets, which may hinder relationships that are only represented at afiner granularity of personality variables.
ATA C OLLECTION
In order to study how personality affects user preferences regard-ing information visualization techniques , we started by choosingwhich contexts we wanted to address (Figure 1): (i) hierarchy, oneof the most common in research (e.g. [36]); (ii) evolution over time,giving the importance of time series data analysis [31]; and (iii)comparison, as it is more appropriate to show differences or sim-ilarities between values at a fixed granularity [4]. We include asimple and familiar scenario with each context in order to stimulate users to reflect on the implications of using each idiom rather thanthe complexity of the data. We focused on minimizing the numberof channels and marks of each graph and keeping them consistentacross contexts, while keeping the same data within a context.Regarding hierarchy, items are all related to each other by theprinciple of containment. We opted for a treemap, a circular packingdiagram, a sunburst, and a Sankey diagram to display the distributionof food consumed by a household within a month. For evolutionover time contexts, we chose line charts with and without points,and area charts. The scenario asked the participant to imagine thatthe data referred to the number of registrants and participants in amarathon held annually in the United States. Finally, we decidedto use radar charts, word clouds, horizontal and vertical bar charts,and pie charts for the comparison context. In particular, the scenariorepresents the levels of the happiness index among six differentcountries (France, Italy, Portugal, Spain, Germany, and the UnitedKingdom).Participants were recruited through standard convenience sam-pling procedures including direct contact and through word of mouth.Our final data set comprises 64 participants (30 males, 34 females)between 18 and 60 years old ( M = . SD = . ) . In addition,they were asked whether they were using glasses or contact lensesand the apparatus used while filling in the questionnaire. Neitherfactor had a significant effect on the experience.Before the experiment, participants were informed about theexperience and invited to agree with a compulsory consent form.They were also informed that they could quit the experiment at anytime. We then collected the FFM five personality traits and its 30facets, and the dimensions of LoC with the Portuguese versionsof the Revised NEO Personality Inventory (NEO PI-R) [7, 19] andthe IPC scale [17, 28], respectively. Afterwards, participants werepresented an online questionnaire which contained a visual exampleof each idiom grouped by context. Participants were firstly promptto read the scenario for the respective context and then assess theirpreference for an idiom by completing a seven-point Likert scaleranging from Low Preference (1) to
High Preference (7). We allowedparticipants to freely change their ratings until they were satisfiedwith all ratings in order to avoid the anchoring bias.
ORRELATION - BASED A NALYSIS
In order to find correlations between personality variables and userpreferences, we used the Spearman’s correlation method (Table 2),as it is preferable when variables feature heavy-tailed distributions orwhen outliers are present [8], and it has been shown as an appropri-ate statistical analysis with Likert scales [23]. Our hypothesis is thata personality dimension from the FFM and/or the LoC is correlatedwith how participants rated their preference for an idiom. Takinginto account the large number of statistical models, we use a Bonfer-roni correction to counteract the problem of multiple comparisons.Therefore, significant p-values are reported at α = . ( r s ( ) = − . , p = . ) , area charts ( r s ( ) = − . , p = . ) and sunburst ( r s ( ) = − . , p = . ) , respectively. In addition,we found that 19 facets showed similarly weak effects. Amongthese facets, we can observe that facets from Agreeableness pointedtowards more effects, followed by Neuroticism and Openness to able 2: Significant results from the Spearman’s correlation tests. Personality Idiom r s p-value N Line Chart with Points -0.267 0.033N1 Treemap -0.364 0.003N3 Line Chart with Points -0.292 0.019N4 Line Chart with Points -0.276 0.027N6 Line Chart with Points -0.277 0.027E1 Sunburst -0.274 0.029E4 Line Chart with Points 0.247 0.049E6 Line Chart without Points -0.283 0.024O Area Chart -0.290 0.020O1 Horizontal Bar Chart -0.269 0.032O2 Line Chart without Points -0.251 0.046O3 Area Chart -0.320 0.010O5 Area Chart -0.340 0.006O6 Sankey Diagram -0.268 0.032A Sunburst -0.285 0.022A2 Sunburst -0.317 0.011A2 Treemap -0.275 0.028A3 Line Chart without Points -0.249 0.047A5 Radar Chart -0.312 0.012A5 Circular Packing 0.249 0.047A6 Sunburst -0.277 0.027A6 Radar Chart -0.263 0.036C3 Line Chart with Points 0.255 0.042C5 Line Chart without Points -0.246 0.050C6 Vertical Bar Chart 0.274 0.028Powerful Others Area Chart 0.320 0.010Powerful Others Line Chart without Points 0.313 0.012Chance Pie Chart 0.382 0.002
Experience. Although both Extraversion and Conscientiousness didnot have an effect, three facets from each of these traits imply aneffect. Regarding LoC, both External dimensions showed weak pos-itive correlations. While Powerful Others may have modelled howparticipants rated both area ( r s ( ) = . , p = . ) and line chartwithout points ( r s ( ) = . , p = . ) , Chance hinted an effecton ratings for the pie chart ( r s ( ) = . , p = . ) . Taking intoaccount the idioms, we can see that line charts suggest the largestnumber of effects related to personality-based user preferences, asmost of Neuroticism, its facets, and facets from Conscientiousnesssuggested correlation effects. At a broader level, evolution over timecontext idioms indicated the largest number of effects (53.57%),followed by hierarchy (28.57%) and then comparison (17.86%). LUSTER - BASED A NALYSIS
Following the work of Sarsam and Al-Samarraie [30], we appliedhierarchical density based clustering [12, 22] to find that the mostappropriate number of clusters to work with was three through sil-houette and DaviesBouldin index scores analysis [27] and Wardscluster method. Then, we used the k-means clustering algorithm [33]to avoid the noise labels that hierarchical density based clusteringyields. We started by normalizing our data and allow the algorithmto run 100 iterations with different centroid seeds using Euclideandistance. The final result contained the best output of 100 consec-utive runs in terms of inertia. As a follow-up, we conducted anANOVA to validate whether each personality trait from one clusterdiffers from the other instances in the other clusters. We found a sig-nificant difference ( p < . ) in between the three clusters regardingNeuroticism, Extraversion, Concientiousness, all dimensions fromLocus of Control, and 18 personality facets out of 30. These resultsshow that all clusters have participants that differ among themselvesin the aforementioned personality variables. Table 3 depicts themeans and standard deviation values for all personality traits of theFFM and dimensions from the LoC. The first cluster ( N = ) no-tably has participants with the highest levels of Conscientiousnessand Internal dimension across clusters. It also includes peoplewith the lowest values on Neuroticism and the External dimen-sions . The remaining traits of Extraversion and Agreeableness showmedium values, while Openness to Experience presents low levels.In contrast, the second cluster ( N = ) shows the lowest valuesfor Conscientiousness . In addition, it features participants with Table 3: Results of the K-means clustering algorithm for each person-ality trait and dimension.
Personality Variable Cluster 1 Cluster 2 Cluster 3M SD M SD M SD
Neuroticism 84.23 20.23 97.18 16.54
Extraversion 112.00 19.02
Chance 15.97 5.87 19.82 5.84
Table 4: Context and preferred idioms with their frequency on toprules for each cluster.
Context Cluster 1 Cluster 2 Cluster 3
Hierarchy
Sunburst (50%) Sunburst (76%) Treemap (37%)
Evolution
Line Chart w/ Points (100%) Line Chart w/out Points (70%) Line Chart w/out Points (74%)
Comparison
Horizontal Bar Chart (71%) Horizontal Bar Chart (71%) Horizontal Bar Chart (71%) the highest values of Extraversion and Agreeableness , while theremaining personality variables show medium values among theclusters. Finally, the third cluster ( N = ) includes participantswith the highest levels on Neuroticism and on both the Externaldimensions . Nevertheless, it presents medium values for Conscien-tiousness and the remaining variables have each the lowest values ofthe set. As we mentioned, it is possible to observe that the trait ofAgreeableness presents similar values across clusters, while Open-ness to Experience, in spite of not showing significant differencesbetween clusters, has very dissimilar values on Cluster 2 comparedto the others. In the real world, Cluster 1 contains people that areorganized and believe in their efforts. In contrast, Cluster 3 includesmoody people that believe the external world has a large influenceover their life. Finally, Cluster 2 contains outgoing and open people.In order to extract information visualization preferences for thedifferent contexts among individuals of those three clusters, we optedfor the Apriori algorithm [13], an association rules method to findcommon patterns. Data preprocessing included the creation of anarray for each participant containing the idiom that they preferred themost for each context. In case of a tie between two or more idiomsin their preference ratings, we included all idioms that tied together.Afterwards, we divided users by their cluster labels and used theApriori algorithm in each cluster. Each run was performed withlower bound minimal values of 0.1 for support, 0.8 for confidence,and 3.1 for lift. An Apriori association rule is often representedas itemA → itemB , which translates into itemB being frequentlypresent in a set of preferences that also contains itemA .We continued our analysis by choosing which rules to focus on theinformation visualization techniques according to the frequency ofeach rule. We started by choosing the rule with the highest frequencyand then choosing rules that had similar item sets until there was norule with common or contradictory associations. Finally, if a contextdid not have a style associated to it, we chose the most frequentpreferred idiom for that context among participants of the cluster,which was the case for the hierarchy context for Cluster 1. Basedon the final set of rules for each cluster, we were able to derivewhich idioms were the most preferred according to the differentcontexts (Table 4). Notably, there are differences in the contexts ofevolution over time and hierarchy. Regarding the evolution over timecontext, both Clusters 1 and 2 prefer a sunburst idiom and Cluster 3participants rate treemaps higher. Compared to the other contexts,the chosen evolution over time idiom was less prominent in Clusters1 (50%) and 3 (36.7%), while Cluster 2 was more consistent intheir preference (76.2%). For hierarchy contexts, while Cluster 1completely prefers line charts with points (100%), the remainingclusters would rather omit the use of those marks. Finally, all clustersstate that an horizontal bar chart is the most preferred idiom to useor comparison data, with frequency values around 71%. This mayhint that the appropriateness of an idiom for a specific problemcontext acts as a stronger regulator compared to personality. ISCUSSION
After analysing our results with both approaches we were able tohave a better understanding of how personality has an effect on in-formation visualization technique preferences. From the correlation-based analysis, results pointed towards effects from the Neuroticism,Openness to Experience, and Agreeableness traits in user preferenceregarding different idioms. Several more facets from all FFM traitsand both External LoC dimensions also suggested a correlation.This lack of significance results is a consequence of the Bonferronicorrection we applied in order to counteract the multiple compar-isons problem. We believe the correlation-based approach is soundfor a smaller number of Spearman correlations, which points ournext step in this research towards the separate analysis of these per-sonality variables to verify whether the results of this study have ahigh false discovery rate. Regarding Neuroticism, the trait itself andthree facets showed a weak negative effect, suggesting that peoplewith higher levels of Neuroticism dislike line charts with points. Infact, we were able to verify the same effect on Cluster 3, whereusers had the highest levels of Neuroticism and they preferred linecharts without points. Additionally, only the cluster with the lowestlevels of Neuroticism (Cluster 1) showed a preference towards linecharts with points. This effect may be given to how people withhigh Neuroticism experience more stress when the idiom containsmore marks, thus more information that may be harder to perceive.Extraversion only showed strong results in the cluster-based ap-proach. While individuals with high and medium levels preferredsunburst as an idiom to represent hierarchy, people with low levelsshowed a preference towards treemaps. Interestingly, individualswith medium levels would rather use line charts with points, contraryto the remaining participants which showed an inclination towardsexcluding points on those charts. This effect may be explained byinteraction effects with the remaining personality variables. In thecorrelation-based approach, three of its facets may have had an effecton preferences for sunburst and line chart idioms, yet all effects wereweak in size.The best clusters produced by the k-means algorithm did notdivide individuals significantly based on Openness to Experience.Nevertheless, we found that independently of the remaining person-ality variables, results suggested it could foster the preference forarea charts. Furthermore, five of its facets hinted negative weakeffects with several idioms, mostly regarding evolution over timeidioms. This is a rather interesting effect, considering how Opennessto Experience has been shown to model how individuals processevolution [1, 9]. Agreeableness was also not significantly differentamong the different clusters, yet, similarly to Openness to Experi-ence, it suggested some significant effects on the correlation-basedapproach along four of its facets. Most of these effects are referringto hierarchy, which may be related to how Agreeableness modelshow people evaluate hierarchical structures of collectivism [29]. Fi-nally, Conscientiousness showed more effects while interacting withthe other personality variables in the cluster-based approach thenby an analysis with correlations. People with high levels tend toprefer line charts with points compared to the remaining population.We believe that this preference may be given to how these peopleprefer an organised approach to life, thus preferring to see idiomswith more detail. We also found that people with high and lowvalues on this trait prefer a sunburst in comparison to a treemap,similar to the Extraversion trait. Concerning the dimensions fromLoC, both External dimensions suggested positive weak correlationeffects. While Powerful Others hinted an effect on the evolution overtime context, Chance did it for the comparison context. In addition,cluster-based analysis showed that people with higher values on these dimensions and the lowest Internal levels among the clustershave a preference for a treemap compared to the sunburst idiom.In contrast, the highest values for Internal and lowest for both theExternal dimensions showed a preference for line charts with points.This effect may be a result of
Internals being faster than
Externals when the former search for items [10] because they use additionalmarks such as points to guide their search.In the light of this, our results suggest that personality is a differ-entiating factor when it comes to designing information visualizationsystems. Looking into our approaches, for example, while facetsfrom Conscientiousness hinted in the correlation-based approach,participants from Cluster 1 preferred line charts with points. Inaddition, results from the Cluster 3 were indicated by the correla-tion results where higher values on Neuroticism or its facets ledparticipants to choose a line chart without points. The same effecthappened with Powerful Others. In contrast, we found dissimilareffects for Extraversion and Agreeableness. Regarding the former,we expected that Cluster 2 would have a preference on the evolutionover time context for line charts with points and on the evolutionover time context different from a sunburst idiom. Moreover, thelatter was not significantly different between clusters. In the light ofthis, we hypothesize that this lack of significance led to an omissionof interaction effects from Agreeableness. In this case, we mustalso address how participants rated each idiom independently ofpersonality. As the Spearman’s correlation effects were all smallin size, we believe that the interaction effects were stronger, as oneindividual perceives information through interactions of all theirpersonality constructs and not only one. Thus, we consider thecluster-based approach to be more appropriated. There are someimportant factors that may explain the lack of significance observedin some of our results. First, since we are tackling a lot of personalityvariables, a larger number of participants would allow conclusionswith a stronger impact in both approaches. In particular, we couldhave a better sampling regarding Openness to Experience, Agree-ableness, and the Internal dimension of LoC. Secondly, althoughthere are more idioms in information visualization, there are moreidioms from these contexts to explore. We also did not control forthe familiarity cognitive bias, which may have had an effect on theresults. Thirdly, the scenario and the complexity of the dataset usedto illustrate the different contexts may have had an effect on howpeople perceived the idioms. Finally, not asking users to performany task rather than rating their preference for the aesthetics of anidiom may not impact visual task analysis.
ONCLUSIONS AND F UTURE W ORK
This exploratory study focuses on personality with two differentpsychological constructs (FFM and LoC) models user preferencesregarding information visualization techniques in three differentcontexts: hierarchy, evolution over time, and comparison. Besidesidentifying which idioms are modelled by personality-based userpreferences, our results suggest important implications that may beused in the design pipeline to customize information visualizationsystems. Future work includes the implementation and testing ofdifferent information visualization systems developed based on ourresults to assess how they affect user preference, performance, ex-perience, and satisfaction. In addition, task types, task complexity,and contexts should be further explored as they may lead to distinctinteractions of users given their individual differences. Finally, weaim to recruit a larger number of participants so that we can exploremore in-depth the personality variables that were not significantlydifferent between clusters. A CKNOWLEDGMENTS
This work was supported by national funds through Fundao paraa Ciłncia e a Tecnologia (FCT) with references UIDB/50021/2020and SFRH/BD/144798/2019.
EFERENCES [1] A. Acerbi, M. Enquist, and S. Ghirlanda. Cultural evolution andindividual development of openness and conservatism.
Proceedings ofthe National Academy of Sciences , 106(45):18931–18935, 2009.[2] M. Braun, R. Chadowitz, and F. Alt. User experience of driver statevisualizations: A look at demographics and personalities. In
IFIPConference on Human-Computer Interaction , pp. 158–176. Springer,2019.[3] E. T. Brown, A. Ottley, H. Zhao, Q. Lin, R. Souvenir, A. Endert, andR. Chang. Finding waldo: Learning about users from their interac-tions.
IEEE Transactions on visualization and computer graphics ,20(12):1663–1672, 2014.[4] D. Cashman, Y. Wu, R. Chang, and A. Ottley. Inferential tasks as adata-rich evaluation method for visualization. In
EVIVA-ML: IEEEVIS Workshop on EValuation of Interactive VisuAl Machine Learningsystems , vol. 7, 2019.[5] C. Conati, G. Carenini, D. Toker, and S. Lall´e. Towards user-adaptiveinformation visualization. In
Twenty-Ninth AAAI Conference on Artifi-cial Intelligence , 2015.[6] P. Costa and R. R. McCrae. The revised neo personality inventory(neo-pi-r).
The SAGE Handbook of Personality Theory and Assessment ,2:179–198, 01 2008. doi: 10.4135/9781849200479.n9[7] P. T. Costa Jr and R. R. McCrae.
The Revised NEO Personality Inven-tory (NEO-PI-R).
Sage Publications, Inc, 2008.[8] J. C. de Winter, S. D. Gosling, and J. Potter. Comparing the pearsonand spearman correlation coefficients across distributions and samplesizes: A tutorial using simulations and empirical data.
Psychologicalmethods , 21(3):273, 2016.[9] G. J. Feist and T. R. Brady. Openness to experience, non-conformity,and the preference for abstract art.
Empirical Studies of the Arts ,22(1):77–89, 2004.[10] T. M. Green and B. Fisher. Towards the personal equation of interac-tion: The impact of personality factors on visual analytics interfaceinteraction. In , pp. 203–210. IEEE, 2010.[11] S. Halko and J. A. Kientz. Personality and persuasive technology: anexploratory study on health-promoting mobile applications. In
Interna-tional conference on persuasive technology , pp. 150–161. Springer, 062010. doi: 10.1007/978-3-642-13226-1 16[12] J. Han, J. Pei, and M. Kamber.
Data mining: concepts and techniques .Elsevier, 2011.[13] M. Ilayaraja and T. Meyyappan. Mining medical data to identifyfrequent diseases using apriori algorithm. In ,pp. 194–199. IEEE, 2013.[14] H.-S. Kim and S. S. Sundar. Motivating contributions to online fo-rums: can locus of control moderate the effects of interface cues?
Health communication , 31:1–13, 09 2015. doi: 10.1080/10410236.2014.981663[15] S. Lall´e and C. Conati. The role of user differences in customization: acase study in personalization for infovis-based content. In
Proceedingsof the 24th International Conference on Intelligent User Interfaces , pp.329–339, 2019.[16] H. M. Lefcourt.
Locus of control: Current trends in theory & research .Psychology Press, 2014.[17] H. Levenson. Multidimensional locus of control in psychiatric patients.
Journal of consulting and clinical psychology , 41(3):397–404, 1973.doi: 10.1037/h0035357[18] H. Levenson. Reliability and validity of the i, p, and c scales-a multidi-mensional view of locus of control. 1973.[19] M. Lima and A. Sim˜oes. Neo-pi-r manual profissional.
Lisboa: CE-GOC , 2000.[20] Z. Liu, R. J. Crouser, and A. Ottley. Survey on individual differencesin visualization. arXiv preprint arXiv:2002.07950 , 2020.[21] R. R. McCrae and P. T. Costa. Self-concept and the stability of person-ality: Cross-sectional comparisons of self-reports and ratings.
Journalof Personality and Social Psychology , 43(6):1282, 1982.[22] L. McInnes, J. Healy, and S. Astels. hdbscan: Hierarchical densitybased clustering.
Journal of Open Source Software , 2(11):205, 2017. [23] G. Norman. Likert scales, levels of measurement and the laws ofstatistics.
Advances in health sciences education , 15(5):625–632, 2010.[24] N. Oscar, S. Mej´ıa, R. Metoyer, and K. Hooker. Towards personalizedvisualization: Information granularity, situation, and personality. In
Proceedings of the 2017 Conference on Designing Interactive Systems ,pp. 811–819, 2017.[25] A. Ottley, R. J. Crouser, C. Ziemkiewicz, and R. Chang. Manipu-lating and controlling for personality effects on visualization tasks.
Information Visualization , 14(3):223–233, 2015.[26] A. Ottley, H. Yang, and R. Chang. Personality as a predictor of userstrategy: How locus of control affects search strategies on tree visu-alizations. In
Proceedings of the 33rd Annual ACM Conference onHuman Factors in Computing Systems , pp. 3251–3254, 2015.[27] S. Petrovic. A comparison between the silhouette index and the davies-bouldin index in labelling ids clusters. In
Proceedings of the 11thNordic Workshop of Secure IT Systems , pp. 53–64, 2006.[28] P. J. P. Queir´os.
Burnout en el trabajo y conyugal en enfermerosportugueses . PhD thesis, Universidad de Extremadura, 2004.[29] A. Realo, J. Allik, and M. Vadi. The hierarchical structure of collec-tivism.
Journal of Research in Personality , 31(1):93–116, 1997.[30] S. M. Sarsam and H. Al-Samarraie. A first look at the effectivenessof personality dimensions in promoting users satisfaction with thesystem.
SAGE Open , 8(2):2158244018769125, 2018. doi: 10.1177/2158244018769125[31] J. Sheidin, J. Lanir, C. Conati, D. Toker, and T. Kuflik. The effect ofuser characteristics in time series visualizations. In
Proceedings ofthe 25th International Conference on Intelligent User Interfaces , pp.380–389, 2020.[32] D. Toker, C. Conati, G. Carenini, and M. Haraty. Towards adaptiveinformation visualization: on the influence of user characteristics. In
International conference on user modeling, adaptation, and personal-ization , pp. 274–285. Springer, 2012.[33] J. Wang, J. Wang, Q. Ke, G. Zeng, and S. Li. Fast approximate k-means via cluster closures. In
Multimedia data mining and analytics ,pp. 373–395. Springer, 2015.[34] C. Ziemkiewicz, R. J. Crouser, A. R. Yauilla, S. L. Su, W. Ribarsky,and R. Chang. How locus of control influences compatibility withvisualization style. In , pp. 81–90. IEEE, 2011.[35] C. Ziemkiewicz and R. Kosara. Preconceptions and individual dif-ferences in understanding visual metaphors. In
Computer GraphicsForum , vol. 28, pp. 911–918. Wiley Online Library, 2009.[36] C. Ziemkiewicz, A. Ottley, R. J. Crouser, A. R. Yauilla, S. L. Su,W. Ribarsky, and R. Chang. How visualization layout relates to lo-cus of control and other personality factors.