Pages

Tuesday, October 31, 2017

ASIST 2017 Panel: Trust and Information Ethics

Trust and Information Ethics Panel

Patient Portals and Human Values 
Rachel N. Simons, Kenneth R. Fleischmann, Yan Zhang, Bo Xie (UT Austin)

Increased adoption of patient portals spurred by pressure from government and insurance companies. Usability cited for lack of patient adoption. UX usually focuses on tasks and requirements, not how values interact with UX.

Online tech that requires patient to have access to technology and have online access. Portals usually tethered to health care institutions's existing EHR. Belief that easier access can better engage patients to manage own health. They don't use interactive functions, only log in once. Human values and ICT. Fleischmann et al 2015 ID six salient values. Their approach to deductive content analysis of user studies. Systematic literature search of user studies of patient portals. 52 articles 2005-2015 all user studies of patient UX with portals, not physician studies, not aggregations. Privacy, security, confidentiality, transparency, trust, agency. Deductive content analysis (Elo & Kyngas, 2008).

Quant results: frequency of values coded in papers--most: security, then transparency, agency, privacy/trust, then confidentiality. Invocation of multiple values only 19% only mentioned one value. Agency and security co-occur the most, followed by security and privacy. In depth qualitative analysis. Thematic analysis. Privacy - negative: logging in in public, positive: the distance afforded by sending online secure messages.

Security and privacy frequently invoked together, focused on technical aspects of design, often conflated. Most documents didn't use actual term "transparency," but it often was reflected in concerns with the patient portal itself vs the transparency of the information presented. Trust in the portal and how it operates vs trust in the health information presented. Other issues included perceptions of and relationships with medical professionals, previous use of a portal, more. "Agency" was often used in a shallow way. No consensus on 'patient empowerment' definition.

Researchers and users do consider human values as part of interactions with patient portals. Users find several values in combination to be salient to their experiences. Qualitative analysis indicates values are rarely articulated or defined by researchers or users. General concepts of usability do not cover values by default.


Human Values and Trust in Scientific Journals, Mainstream Media, and Fake News
Nitin Verma, Kenneth R. Fleischmann, Kolina S. Koltai (UT Austin)

Media habits have been changing dramatically. Tremendous growth in fake news sharing on social media. Americans express only moderate trust in most news sources. Trust in institutions declining over time. Specifically for science news, majority of Americans get it from mainstream media moreso than from magazines, documentaries, and museums. But trust mainstream media least to report science news. Only have of social media users click through to follow up on sources. Degree to which evidence provided in scientific coverage by various outlets (Infographic - American Council on Science and Health Image/RealClear Science).

Human values - what is it that motivates peoples' behavior? Schwartz Value Inventory from social psychology used. Does including hyperlinks influence trust, do values of reader influence trust? Does the type of hyperlink influence trust? Used Amazon's mechanical Turk (MTurk), used to apply to studies in trust in online information. Used no retweets or replies and used new egg photo on Twitter to remove confounding factors. 21 Item PVQ Portrait Value Questionnaire by Schwartz. Given 10 tweets in a timeline to view with either no link, hidden link, direct link. Each shown an equal amount of times shown. Measured to see if clicked on links, at what point, and asked to rate trust. Post survey not included in this presentation. 205 HITs, Nonparametric methods because of ordinal rating scales and fewer assumptions needed to be made about underlying distributions. Posts with hyperlinks garnered significant level of more trust. (p<0 -="" .001="" 5="" across="" all="" and="" as="" better.="" better="" clicked="" condition.="" cues="" each="" fake="" for="" hidden="" high="" hyperlinks="" in="" indeed="" journal="" journals="" link="" links="" mainstream="" media="" msm="" nbsp="" news.="" news="" not="" of="" p="" perform="" performed="" post="" ratings="" scientific="" significantly="" sources="" terms="" than="" trust="" trustworthiness.="" types.="" were="" when="">
If links not clicked, what factors might be affecting how users express trust? Clicking through overrides personal values. Values play more important role in absence of other cues. Does including hyperlinks influence trust? Yes, increases trust. Does type increase trust? Yes. Do values of reader influence trust? Yes.


Trust in Qualitative Data Repositories
Rebecca D. Frank, University of Michigan

Part of larger project at UM, an IMLS funded project Elizabeth Yakel is PI.

Qualitative data in the field of education: talking abut educational records of practice and what that looks like is video of children and teachers in K-12 classrooms, video may be re-used, includes lesson plans, worksheets, student work, student assignments, demographic info about classrooms, schools districts. This is particularly interested in video. Difficult and costly to produce this kind of data in terms of cost of equipment, skill to capture dynamic aspect of classroom, permission and consent issues, K-12 not usually open and visible so cultural issues make capturing this data sensitive. Making available for sharing and reuse for others, those issues are compounded by technical issues of sharing large vide files, confidentiality, people depicted in or producing videos members of same professional community of practice as subjects. So, not a strong culture of dat sharing in education when it comes to video records of practice. But it is becoming more common. Repositories: ATLASn part of national board for professional teaching standards), TIMSSVideo (one dataset), Teaching & Learning Exploratory (Umich), Measures of Effective Teaching. Usually repositories spring up when initial goal was sharing the data. Usually a pricing model which is highly variable.

Previous research says: they're designed for tech capabilities and not user needs. Specialized tools repositories provide are a barrier and not a facilitator. "Trust is the data reuser's belief that using it will lead to positive outcomes, leading to reuse (Yoon, 2016); Bak 2016 Trust is social and contingent. Pirson & Malhotra (2011)- image: Theoretical framework of organizational trust. (Organization Science 22.4).

For this model, depth of interaction is extent or intensity of data reuse, and locus is relationship to repository. Tons of RQs: Number and range of repository data reusers interact with, depth of interactions between data reusers and repository staff, [more]. Survey (139, only 17 reported interaction with repository staff) and semistructured interviews (44; 21 interacted with staff). Found interaction with staff didn't always occur at point of data reuse. People who did interact with staff included incidental interaction at conferences, asking for help, fall along spectrum of internal and external.

Most useful are curated (positive), (negative) no responsiveness to taggers probably meant even worse response for users.

Trust: Data quality (are there problems with metadata and description?); co-production of data between repository and producer (work of original producer reliable?); responsiveness of staff (even if problem isn't completely solved, more overall positive view); curation process was strongest link to repository trust (more transparent, the more likely respondents were to find them trustworthy).

Participants didn't fall into the boxes of the original model, but a spectrum. For repositories: opportunities for interaction and participation in metadata creation, transparent data curation practices, responsiveness of repository staff.





No comments: