Human-Centered Responsible AI: Current & Future Trends
A report from the special interest group at the ACM CHI Conference on Human Factors in Computing Systems 2023.
In April 2023, I co-organized a Special Interest Group (SIG) in the most prestigious Human-Computer Interaction (HCI) conference in Hamburg in 2023 (about 4000 in-person total participants). Together with 15 world-known experts (e.g., from Microsoft, Google, Meta, IBM, Bell Labs, and academics from prestigious universities) in Responsible AI (RAI), we ran a session with 100+ in-person attendees. You can find the proposal for the SIG here.
During the session, participants were split into groups and discussed current and future trends in responsible AI. Each group had to develop a situation, complication, and proposal slide related to responsible AI. In total, we had 12 of these slides. The resulting proposals could give us ideas about future projects and where the community is moving. Two sample topics that emerged from the groups are:
Fairness, power dynamics, biases, diversity & beyond:
In the ever-evolving landscape of AI, one of the most perplexing challenges we face is determining accountability and responsibility. It’s a complex issue that becomes even murkier when considering the legal aspects. With the law struggling to keep up with the rapid advancements in technology, we find ourselves in a gray area where the rules are yet to catch up with the tools. Crafting effective laws takes time, and the cultural contexts across different countries further complicate matters.
Moreover, there is a constant tug-of-war between protecting business interests, user rights, and the rights of creators. Therefore, there is a need to distinguish between legal, ethical, and socio-cultural aspects of fairness. We should also consider the multiple roles and contexts within AI systems, where users can be data sources, decision targets, or artifact recipients. We must navigate these complexities thoughtfully and deliberately to shape a future where AI aligns harmoniously with our values and aspirations.
Theories of HCI and their applications in AI research:
The absence of established theories in RAI poses a challenge. Therefore, understanding the role of HCI theories and their applications RAI is essential. Psychological theories can inform design and shed light on technology’s impact, while normative lenses and process frameworks provide ethical and practical guidance. Integrating theories into RAI practices involves unpacking RAI stages and developing RAI-native theories. This approach enables responsible decision-making and anticipates ethical considerations, propelling the advancement of RAI.
Learn more about the other future trends in RAI here.
Here is the complete list of all organizers:
🤖 Mohammad Tahaei, Nokia Bell Labs
🤖 Marios Constantinides, Nokia Bell Labs
🤖 Daniele Quercia, Nokia Bell Labs
🤖 Sean Kennedy, Nokia Bell Labs
🤖 Michael Muller, IBM Research AI
🤖 Simone Stumpf, University of Glasgow
🤖 Q. Vera Liao, Microsoft Research
🤖 Ricardo Baeza-Yates, Northeastern University
🤖 Lora Aroyo, Google
🤖 Jess Holbrook, Meta
🤖 Ewa Luger, University of Edinburgh
🤖 Michael Madaio, Google
🤖 Ilana Golbin Blumenfeld, PwC
🤖 Maria De-Arteaga, University of Texas at Austin
🤖 Jessica Vitak, University of Maryland, College Park
🤖 Alexandra Olteanu, Microsoft Research