Javascript must be enabled to continue!
Towards Agency-Enhancing Artificial Intelligence

As AI systems become increasingly capable, a troubling pattern emerges. The dominant paradigm optimizes for automation, replacing human judgment, skill, and decision-making, rather than augmenting and enhancing it. This talk argues that this is not an inevitable feature of technological progress, but a design and values choice, and one we can and should reverse.

Drawing on research at the intersection of HCI, AI Ethics and Safety, this talk advances the idea of agency-enhancing AI: systems designed not to accomplish tasks on behalf of people, but rather to expand the range of what people can understand, decide, and do for themselves. I argue that alignment, as a concept, is itself part of the problem. Vague by design and reductive by consequence, alignment presupposes a stable, singular set of human values that AI can be tuned toward. But human needs are plural, contested, and contextual, shifting across cultures, communities, and moments in time. No optimization target captures that complexity, and no technical framework resolves the political question of whose values get encoded. Rather than better alignment, what we need is AI systems designed around a richer understanding of human agency, one that embraces multiplicity rather than flattening it.

Grounding this argument in empirical work spanning language model bias, human-AI interaction design, and the role of AI in education and knowledge work, the talk will explore what it means to put human agency at the center of AI development: who benefits from agency-enhancing design, how we evaluate it, and what institutional and technical changes are needed to make it the norm rather than the exception.

Keynote
Orestis Papakyriakopoulos

Back

ACM GreekCHI-CYCHI Symposium 2026


We are excited to invite you to the ACM GreekCHI-CYCHI Symposium 2026, held on November 6-7, 2026 jointly in Athens, Thessaloniki, Nicosia and Online.