AI & Society最新文献

Social care robots (SCR) have come to the forefront of the ethical debate. While the possibility of robots helping us tackle the global care crisis is promising for some, others have raised concerns about the adequacy of AI-driven technologies for the ethically complex world of care. The robots do not seem able to provide the comprehensive care many people demand and deserve, at least they do not seem able to engage in humane, emotion-laden and significant care relationships. In this article, we will propose to focus the debate on a particularly relevant aspect of care: trust. We will argue that, to answer the question of whether SCR are ethically acceptable, we must first address another question, namely, whether they are trustworthy. To this end, we propose a three-level model of trust analysis: rational, motivational and personal or intimate. We will argue that some relevant forms of caregiving (especially care for highly dependent persons) require a very personal or intimate type of care that distinguishes it from other contexts. Nevertheless, this is not the only type of trust happening in care spaces. We will adduce that, while we cannot have intimate or highly personal relationships with robots, they are trustworthy at the rational and thin motivational level. The fact that robots cannot engage in some (personal) aspects of care does not mean that they cannot be useful in care contexts. We will defend that critical approaches to trusting SCR have been sustained by two misconceptions and propose a new model for analyzing their moral acceptability: sociotechnical trust in teams of humans and robots.

This manuscript presents the design and implementation of a chatbot integrated with moral foundations, to explore its influence on addressing divisions in democratic societies. The study uses generative AI to simulate conversations that embody specific moral viewpoints, categorized into Individualizing and Binding foundations from Moral Foundations Theory. These chatbots engage participants in discussions on an ideologically sensitive topic, nuclear abolition, to observe the impact on communication dynamics and opinion formation. The experiment, conducted with participants in Japan, shows that the chatbots effectively conveyed their designed moral foundations and influenced how participants perceived and interacted with them. Only a limited number of participants were able to identify the partner as an AI. The Individualizing chatbot, which emphasizes individual rights and welfare, were associated with more positive interpersonal impressions than the Binding chatbot, which emphasizes group cohesion and social order. Furthermore, the study shows that discussions with these chatbots can significantly change participants’ opinions on the issues under discussion, demonstrating the potential for and cautions against AI-driven tools in ideological intervention and moral education. The findings argue for the careful development and use of AI in addressing societal divides, highlighting both the potential benefits and ethical challenges.

Explainable artificial intelligence (XAI) aims to provide insights into the logic of automated decisions with the goal of promoting fairer, more transparent, and more trustworthy automated decision-making. Despite mounting regulatory pressure, changing consumer expectations, and a growing stream of XAI-related research, few consumer-facing applications of XAI exist. In anticipation of future XAI-enabled products and services, we use ethical foresight analysis to investigate the possible consequences of monetizing explanations. By developing a conceptual artifact we call an explanation platform, we analyze what could happen when digital advertising is fused with XAI. We explore the platform’s business and design logic, examine its potential social and ethical impact, and describe several plausible explanation manipulation scenarios and strategies. We find that while XAI monetization could incentivize industry adoption of XAI technology and expand algorithmic recourse across society, it could also lead to corrupted forms of explanations optimized for profit-driven objectives. Overall, our foresight analysis makes the case for the economic and technological feasibility of monetized XAI, but raises concerns about its desirability in liberal democratic societies.

Judgments of misinformation are made relative to the informational preferences of the communities making them. However, informational standards change over time, inducing distribution shifts that threaten the adequacy of machine learning models of misinformation. After articulating five kinds of distribution shifts, three solutions for enhancing success are discussed: larger static training sets, social engineering, and dynamic sampling. I argue that given the idiosyncratic ontology of misinformation, the first option is inadequate, the second is unethical, and thus the third is superior. However, I conclude that the prospects for machine learning models of misinformation are far weaker than most have presupposed, given that both epistemic and non-epistemic values are difficult to operationalize dynamically in machine code, rendering them surprisingly at most a species of recommender systems rather than literal truth detectors.

A catastrophic AI threat model is a rigorous exploration of some particular mechanisms by which AI could potentially lead to catastrophic outcomes. In this article, I explore a polycrisis threat model. According to this model, AI will lead to a series of harms like disinformation and increased concentration of wealth and power. Interactions between these different harms will make things worse than they would have been had each harm operated in isolation. And the interacting harms will ultimately cause or constitute a catastrophe. My aim in this paper is not to defend the inevitability of such a polycrisis occurring. Instead, I aspire merely to establish that polycrisis-driven catastrophe is sufficiently plausible that it calls for further exploration. In doing so, I hope to emphasise that alongside worries about AI takeover, those concerned about catastrophic risk from AI should also take seriously worries about extreme power concentration and systemic disempowerment of humanity.

Amid the pervasive integration of AI technologies across societal and industrial domains, understanding users’ trust in these systems becomes increasingly crucial. This study addresses the growing need to understand users’ trust in Generative Artificial Intelligence (GenAI) and explores the societal implications of this type of trust. Based on the socio-technical systems theory, this work employs the FAT (Fairness, Accountability, Transparency) framework and humanness factors of AI, anthropomorphism, social presence, and emotions, as antecedents of users’ human-like trust, which is proposed to influence users’ attitudes, perceived performance, and behavioral intentions. Structural equation modeling analysis (N = 244) reveals that fairness significantly enhances trust, while accountability and transparency do not. Social presence and emotions positively impact trust, whereas anthropomorphism shows no significant effect. Furthermore, trust shapes users’ attitudes, perceived performance, and behavioral intentions toward GenAI systems. This study contributes to the AI adoption and user trust literature by illuminating the main antecedents of human-like trust and showing its impact on user acceptance from a social-technical perspective. Beyond the academic contribution, this research highlights the broader societal relevance of user trust in GenAI, particularly regarding public concerns over black box issues and humanness features of GenAI systems.

The rise of ChatGPT has emphasized the need for an improved conceptual understanding of users’ agency when interacting with artificial intelligence (AI) systems for healthcare. Australian ChatGPT users (N = 216) completed a repeated measures online survey. Hierarchical regression analyses assessed the influence of demographic factors (age and gender), Technology Acceptance Model constructs (perceived usefulness and perceived ease of use), and extended variables (trust, privacy concerns) on users' behavioral intentions to use ChatGPT for physical and mental healthcare. The proposed model was partially supported: the findings emphasized the need to establish user trust in ChatGPT and its perceived usefulness in both areas of healthcare. Privacy concerns were a significant predictor of intentions to use ChatGPT for mental healthcare with perceived ease of use predicting intentions to use ChatGPT for physical healthcare. The findings indicate predictors of uses of AI cannot be generalized across healthcare types and unique drivers should be considered.

Despite significant advancements in AI and automated driving, a robust ethical framework for AV decision-making remains undeveloped. Such a framework requires clearly defined moral attributes to guide AVs in evaluating complex and ethically sensitive scenarios. Existing frameworks often rely on a single normative ethical theory, limiting their ability to address the nuanced nature of human decision-making and leading to conflicting outcomes. Augmented Utilitarianism (AU) offers a promising alternative by integrating elements of virtue ethics, deontology, and consequentialism into a non-normative framework. Grounded in moral psychology and neuroscience, AU employs mathematical ethical goal functions to capture societally aligned attributes. In this study, we propose and evaluate a method to elicit these attributes for AV decision-making. One hundred participants were presented with traffic scenarios, including critical and non-critical situations, and tasked with evaluating the relevance of an initial set of 11 attributes (e.g., physical harm, psychological harm, and moral responsibility) while suggesting additional relevant attributes. Results identified two new attributes—environmental harm and energy efficiency—and revealed that four attributes (physical harm, psychological harm, legality of the AV, and self-preservation) varied significantly between critical and non-critical scenarios. These findings suggest that the weight of attributes in ethical goal functions may need to adapt to situational criticality. The method was validated based on key evaluation criteria: it demonstrated sensitivity by producing varying relevance scores for attributes, was deemed relevant by participants for eliciting AV decision-making attributes, and allowed for the identification of additional attributes, enhancing the robustness of the framework. This work contributes to the development of a dynamic and context-sensitive ethical framework for AV decision-making.