Trust in robot-mediated health information
aa r X i v : . [ c s . H C ] D ec Trust in robot-mediated health information*
David Cameron , Marina Sarda Gou, and Laura Sbaffi Abstract — This paper outlines a social robot platform forproviding health information. In comparison with previousfindings for accessing information online, the use of a socialrobot may affect which factors users consider important whenevaluating the trustworthiness of health information provided.
I. INTRODUCTIONAhead of the use of social robotics healthcare and healthinformation contexts, it is important to identify what prospec-tive health information users would want from such services.Technology is advancing rapidly and health information isincreasingly mediated by computers and accessed online[1]; there may be valuable insights from the current useof online health information to be considered in the furtherintroduction of social robotics in health care.Although online websites and social robots have poten-tial to deliver accurate health information, the interactionexperience for each of these may be quite different. Incomparison with virtual agents, robots are rated as havingmore sociability, responsiveness, and trustworthiness [2]; adynamic and responsive agent could also be seen as a socialentity and treated as if it was a real person [3]. There isalready some evidence of these differences in the contextsof educational information [4]. Thus, people seeking healthinformation might not have the same reactions interactingwith static information online in comparison to a social robot;this may affect how they evaluate information provided.In the present article, we specifically consider the issue oftrust. Trust in robots is still an emerging area in the fieldof Human-Robot Interaction (HRI) [5] but nonetheless isincreasingly recognised as influencing the course of inter-action, such as people’s willingness to follow advice [6].Trust in HRI has a contextual basis [7], so the nature of theinteraction - in this case receiving health information - mayspecifically frame people’s trust towards a social robot inways that resemble other health information contexts (e.g.,from a pharmacist or from health websites).
A. Trusting Health Information
Understanding how people come to trust the informationand advice they find online has been an important issuesince the widespread adoption of digital technologies [8].Despite the introduction of standards, concerns over infor-mation quality, accuracy and credibility are still echoed by *This work was funded by The University of Sheffield, InformationSchool’s researcher development fund Information School, The University of Sheffield, Sheffield, UK [d.s.cameron] [l.sbaffi]@sheffield.ac.uk Department of Psychology, The University of Sheffield, Sheffield, UK [email protected] researchers examining the provision of health informationmaterial across a range of conditions [9]. Recent research hasdemonstrated that there are many factors to consider whenaddressing trust formation in health information, all of whichare interdependent and interconnected [1].
B. Research Aim
This exploratory work examines if the use of a social robotplatform influences which factors users consider important inevaluating the trustworthiness of health information offered.Outcomes in this study are compared against previous find-ings for users searching for health information online [1].II. M
ETHOD
A brief, staged, video scenario was shown to represent apotential use of robots in healthcare. In the video, Pepper firstintroduced itself as a source of health information. A personthen asks Pepper for advice on how to treat a minor injury(knee pain) and Pepper delivers advice through synthesisedspeech. All advice came from the relevant NHS website.Trust questions were drawn from a scale used to establishtrust in online health information [1]. Questions were adaptedto reflect the use of a social robot as an interface for healthinformation (e.g., ‘The extent to which the article gives meinformation that I can use’ became ‘The extent to which the robot gives me information that I can use’).As with the original study [1], questions were presented asa bank of 5-point Likert-style statements to investigate par-ticipants’ reports of the relative importance of each statementin determining a health-information robot’s trustworthiness.Participants were recruited through the University ofSheffield volunteers list. 97 participants (34 identified asmale, 63 as female; Mean age = 32.78) passed the atten-tion screening test (accurately describing the health issuepresented in the video) and completed the questionnaire.Participants were invited to take part in a prize draw asrecompense for their time.III. R
ESULTS
Responses to the questions were collated across the 8 mainfactors previously identified by students for trusting onlinehealth information [1]. Mean and SD scores for these factorsare reported in Table 1 and factors are ranked by their meanimportance.
A. Comparison with historical data
The ranked importance of factors for the current resultsdiffers substantially from historical data for the factor rank-ings for online health information. Table 2 highlights the
ABLE I
FACTORS RANKED BY REPORTED IMPORTANCE factor factor Definition Mean SDStyle The way informationis presented and written 4.30 .72Credibility Information is believableand impartial. 4.29 .73Ease of use Ease of accessing andusing the robot 4.20 .72Content Information’s characteristics:reliability, accuracy, currency 4.15 .82Usefulness Extent the user can makeuse of the information 4.09 .83Verification Apparent expertise andconsistency 3.86 .94Recommendation Recommendations fromknown persons 3.26 1.11Brand Robot brand indicatorsand reputation 3.04 1.03 relative importance in terms of rank for each factor used indetermining the trustworthiness of health information onlineor delivered by a robot platform.
TABLE II
FACTOR RANKINGS COMPARED WITH HISTORICAL DATA
Historical Ranking [1] Current Ranking ChangeCredibility Style ↑ Content Credibility ↓ Style Ease of use ↑ Usefulness Content ↓ Brand Usefulness ↓ Verification Verification –Ease of use Recommendation ↑ Recommendation Brand ↓ IV. D
ISCUSSION
The results indicate that the use of a social robot as aplatform for health information may influence people’s viewson what to consider when it comes to assessing the trust-worthiness of the information. Specifically, in comparison tohistorical data for people assessing online delivery of healthinformation [1], robot-mediated information may result inpeople prioritising different aspects of such information.In the current work, style is recognised as the mostimportant aspect of assessing trustworthiness. Given theunfamiliarity most people would have in interacting witha social robot compared to their experience of using sitesonline, this may reflect clarity as an important issue in(at least initial) interaction [10]. Similarly, ease of use isranked as higher importance in the current study than inthe previous. Again, when facing new health technology,ease of use is a significant predictor in its acceptance [11];the novelty of social robots in this context could potentiallyinflate the importance of ease of use.In contrast, a robot’s branding is ranked as least importantfor determining trustworthiness. Without established repu-tations to associate with different robots or manufacturerbrands, a robot’s brand may provide very little information tousers; this may present an opportunity for health informationproviders to attach their own brands to such devices.
A. Limitations
The are distinct differences in focus and methodologybetween this and the prior study [1]; caution is advisedwhen interpreting these early results. The previous workcomprised participants considering their own active searchfor health information on existing technology, whereas thisstudy asks participants to observe a novel interaction andimagine. Nonetheless, the results point to potentially viableareas for further research.
B. Future directions
Of note, credibility is ranked high across both online androbot-mediated health information; trust in these health con-texts may have a significant social element - the informationis perceived to be provided with the user’s interest and well-being in mind - as seen in face to face healthcare [12]. Givenusers may perceive robots as social agents, and the specificsocial contexts in which health information may be accessed,it may be productive to further explore this factor in termsof trust in robots as a social construct.R
EFERENCES[1] J. Rowley, F. Johnson, and L. Sbaffi, “Students’ trust judgements inonline health information seeking,”
Health informatics journal , vol. 21,no. 4, pp. 316–327, 2015.[2] A. Powers, S. Kiesler, S. Fussell, and C. Torrey, “Comparing a com-puter agent with a humanoid robot,” in
Proceedings of the ACM/IEEEinternational conference on Human-robot interaction , 2007, pp. 145–152.[3] B. Reeves and C. I. Nass,
The media equation: How people treatcomputers, television, and new media like real people and places .Cambridge university press, 1996.[4] S. H. Hsu, C. Y. Chou, F. C. Chen, Y. K. Wang, and T. W. Chan,“An investigation of the differences between robot and virtual learningcompanions’ influences on students’ engagement,” in
First IEEEInternational Workshop on Digital Game and Intelligent Toy EnhancedLearning (DIGITEL’07) . IEEE, 2007, pp. 41–48.[5] K. E. Schaefer, “Measuring trust in human robot interactions: De-velopment of the “trust perception scale-hri”,” in
Robust Intelligenceand Trust in Autonomous Systems . Springer, Boston, MA., 2016, pp.191–218.[6] A. Freedy, E. DeVisser, G. Weltman, and N. Coeyman, “Measurementof trust in human-robot collaboration,” in
International Symposium onCollaborative Technologies and Systems . IEEE, 2007, pp. 106–114.[7] D. Cameron, J. Aitken, E. Collins, L. Boorman, A. Chua, S. Fernando,O. McAree, U. Martinez Hernandez, and J. Law, “Framing factors:The importance of context and the individual in understanding trust inhuman-robot interaction,” in
International Conference on IntelligentRobots and Systems . Hamburg, Germany, 2015.[8] E. Sillence, P. Briggs, P. R. Harris, and L. Fishwick, “How do patientsevaluate and make use of online health information?”
Social science& medicine , vol. 64, no. 9, pp. 1853–1862, 2007.[9] L. Daraz, A. S. Morrow, O. J. Ponce, W. Farah, A. Katabi, A. Majzoub,M. O. Seisa, R. Benkhadra, M. Alsawas, P. Larry, et al. , “Readabilityof online health information: a meta-narrative systematic review,”
American Journal of Medical Quality , vol. 33, pp. 487–492, 2018.[10] A. Rossi, P. Holthaus, K. Dautenhahn, K. L. Koay, and M. L. Walters,“Getting to know pepper: Effects of people’s awareness of a robot’scapabilities on their trust in the robot,” in
Proceedings of the 6thInternational Conference on Human-Agent Interaction , 2018, pp. 246–252.[11] F.-Y. Pai and K.-I. Huang, “Applying the technology acceptance modelto the introduction of healthcare information systems,”
TechnologicalForecasting and Social Change , vol. 78, no. 4, pp. 650–660, 2011.[12] J. E. Croker, D. R. Swancutt, M. J. Roberts, G. A. Abel, M. Roland,and J. L. Campbell, “Factors affecting patients’ trust and confidencein gps: evidence from the english national gp patient survey,”