R. L. Rabb Symposium on Embedding AI in Society Summary
By Erin Morrow and Veljko Dubljević
|Image courtesy of Pixabay
Joanna Bryson commenced with a talk entitled “Bias, Trust, and Doing Good: Scientific Explorations of Topics in AI Ethics.” A significant portion of the presentation explored how trust is operationalized, with Bryson arguing that trust is necessary for interactions in which the outcome is not certain. In particular, she highlighted experiments exploring cross-cultural differences in social decision-making when faced by uncertainty. Cultural contexts appeared to impact how inclined a given population was to make prosocial or antisocial choices—one insight being that, in an economic game of trust, participants from Boston in the United States preferred to “[exploit] advantages of the present system” rather than take economic punishments into their own hands. In contrast, participants from cities such as Muscat, Oman and Athens, Greece preferred to intervene, opting to use a mix of prosocial and antisocial strategies (i.e., punishing free-riders, who contributed less to group wealth than the punisher, and punishing those who contributed as much as or more than the punisher, respectively). This variability of social tendencies, then, impacts how people interact with entities (presumably including AI) in encounters that require trust. Different regions and peoples will likely experience different relationships with AI, as we do with each other.
Frank Pasquale followed with a discussion centered on “Renewing the Political Economy of Automation.” He took on the popularized notion that—e.g., if the diagnosis accuracy of medical AI continues to improve—AI will overtake the contribution of physicians in the practice of medicine. He acknowledged that, while AI will likely assist in helping provide more accurate diagnoses, the doctor’s role implicates qualities beyond mere pattern-driven model execution. Take the bias present in datasets with a disproportionate lack of dark skin representation in dermatology: if exclusively left to AI, cases of skin cancer in these populations may go undetected. Pasquale instead suggested a ‘melting pot’ scenario in which AI serves a complement to physicians rather than a substitute. More broadly, he predicted that certain human labor will actually become more valued as lower-paying, more dangerous, and more tedious jobs will be occupied by AI. Therefore, he argued, societal and legal discussion must take place regarding what the public values most, as well as what role AI should therefore serve in the workforce.
|Image courtesy of Future Atlas on Flickr
Offering more of an anthropological perspective, Benjamin Kuipers spoke as the symposium neared its conclusion. His presentation on “Hunting for Unknown Unknowns: AI and Ethics in Society” found parallels in Bryson’s earlier talk as he addressed what he claimed to be the fundamental pillar of civilization: trust. He posited that trust is an essential contributor to cooperation and social norms, with the evolutionary benefit of increasing the resources available to society. However, Kuipers suggested that AI has the potential to threaten trust. All AI systems use models, which are inherent simplifications of reality. This leads to a failure in accounting for “unknown unknowns,” or factors which cannot be predicted from the model. In particular, the common model of utility maximization can damage trust when not used thoughtfully (i.e., when this maximization takes advantage of vulnerability and thus dampens cooperative actions) and/or when the oft-tricky utility measure turns out to be inappropriate (e.g., a function that merely raises financial value may not result in desired outcomes). Until the influence of AI on trust can be more precisely evaluated, Kuipers argued, society has work to do.
|Image courtesy of Brian J. Matis on Flickr
This programming, along with an abstract spotlight each morning, offered a multitude of disciplinary perspectives on the social situation of AI. Prominent themes included AI’s potential influence on or even jeopardization of interpersonal and societal trust, different forms of AI bias and how they might be addressed, and synergistic relationships with AI in the economy and job market. Interwoven in these discussions were questions of responsibility, accountability and oversight, and human-AI dynamics. The topics featured also shared themes in neuroscience, touching on the ethical implications of intelligent systems and devices designed to be ‘brain-like’ (e.g., artificial neural networks). All in all, the symposium served as a vessel for important dialogues about our future with AI, which look to be extended in a special issue on Embedding AI in Society in the journal AI & Society: Journal of Culture, Knowledge and Communication.
The AI in Society Group at NC State is actively seeking contributions: abstracts and manuscripts, as well as any inquiries, can be submitted to [email protected].
Want to cite this post?
Morrow, E. & Dubljević, V. (2021). R. L. Rabb Symposium on Embedding AI in Society Summary. The Neuroethics Blog. Retrieved on , from http://www.theneuroethicsblog.com/2021/05/r-l-rabb-symposium-on-embedding-ai-in.html