Social Good and Ill: Implications for Research, Practice, and Policy

Permanent URI for this collection


Recent Submissions

Now showing 1 - 6 of 6
  • Item
    A Framework for Socio-Developmental Ethics in Educational AI
    ( 2023-01-03) Tuomi, Ilkka
    In recent years there have been many attempts to create ethical frameworks for AI. Theoretical concepts, such as privacy, fairness, transparency, explainability, responsibility, risk, and trustworthiness have been used as key elements in these frameworks. The use of these concepts is often justified by their wide use in similar frameworks and guidelines but does not seem to result from any coherent shared theoretical foundation. Educational and developmental theories and research have so far had little impact on ethical debates but become important when AI is used in education and learning (AIEd). A socio-developmental view on ethics naturally emerges in the educational context, and the paper shows that it has important implications also beyond the education sector. This paper describes an ethical framework structured in three thematic domains: agency, social fairness, and justified choice, that links AI with theories of education and human development, opening new ways to understand ethics of AI and the social and technical challenges and opportunities in AIEd.
  • Item
  • Item
    Teaching Case Study: Gender Data Trouble in a Student Information System
    ( 2023-01-03) Stelmaszak, Marta ; Wagner, Erica
    In 2022, StateU, a large public university in the United States, embarked on a project to collect and use personal pronouns in its information systems. The project lead and functional expert was StateU's Administrative Leader. As she prepared for the first project meeting, she reflected on lessons learned from a past project she led to expand the collection of student gender data to record legal sex, gender identity, and sexual orientation. That project involved navigating challenging decisions about user interface design, underlying databases, data privacy and security, and reporting, underpinned by the desire to best serve minoritized and vulnerable populations. She recalled that: "A society with more data about LGBTQ people is not automatically a society that is better for LGBTQ people". She wondered if collecting pronoun data was the right choice in the first place.
  • Item
    Using Instrumental Mechanisms to Support Humanistic Goals: The Case of Two Intelligent Personal Assistants
    ( 2023-01-03) Cranefield, Jocelyn ; Doyle, Cathal ; Ekandjo, Talitakuum
    Calls have been made for information systems to go beyond supporting the instrumental outcomes traditionally associated with business imperatives to foster more humanistic outcomes. This study explores the mechanisms used by two intelligent personal assistants (IPAs) to promote humanistic goals such as pro-social behaviour. We identify four key mechanisms through which the IPAs support humanistic goals and draw on humanistic management literature to identify the humanistic goals supported. The mechanisms are (1) humanistic framing of analytics and goals, (2) persuasion, (3) automation of humanistic actions, and (4) anchoring humanistic goals to instrumental outcomes. The study raises issues about the moral implications of instrumentalising humanistic outcomes and suggests a need for theory to understand the role of Human-AI interaction in promoting humanistic outcomes. We propose a need for investigations into how and whether human-AI interactions can foster authentic humanistic outcomes in practice.
  • Item
    Innovating within Institutional Voids: A Digital Health Platform in India
    ( 2023-01-03) Ahuja, Suchit ; Chan, Yolande ; Sadreddin, Arman
    Most of the literature on digital innovation assumes availability of resources and access to markets and intermediaries. Institutional voids – lack of formal and informal arrangements – are generally seen as detrimental to digital innovation. While the extant literature provides insights about how some innovation can take place within institutional voids, it largely ignores the role of digital platforms. Based on field work in India, we examine how digital platforms can interface with institutional voids to create social and economic impacts. We find that platforms can address socio-economic challenges by framing, aggregating, and networking within institutional voids. Using an illustrative case study in rural India, where voids and constraints are prevalent, our research highlights how platforms can take strategic actions to develop socio-digital solutions to serve marginalized populations while earning sustainable revenues. We highlight dynamic interactions among physical, social, and digital layers that help platforms reframe constraints and address institutional voids.
  • Item
    Text Versus Paratext: Understanding Individuals’ Accuracy in Assessing Online Information
    ( 2023-01-03) Suntwal, Sandeep ; Brown, Sue
    Fake news has emerged as a significant problem for society. Recent research has shown that shifting attention to accuracy improves the quality of content shared by individuals, thereby helping us mitigate the harmful effects of fake news. However, the parts of a news story that can influence individuals’ ability to discern the true state of information presented to them are understudied. We conducted an online experiment (N=408) to determine how different elements (text and paratext) of a news story influence individuals’ ability to detect the true state of the information presented. The participants were presented with the headline (control), main text, graphs/images, and sharing statistics of true and fake news stories and asked to evaluate the story's accuracy based on each of these elements separately. Our findings indicate that individuals were less accurate when identifying fake news from headlines, text, and graphs/images. When asked to evaluate the story based on sharing statistics, they were able to distinguish fake stories from real news with greater accuracy. Our findings also indicate that heuristics that apply to true news are ineffective for detecting the veracity of fake news.