AIES-18: The Genesis of AI Ethics as a Formal Discipline

TL;DR

The inaugural AAAI/ACM Conference on AI, Ethics, and Society (AIES-18) marked a pivotal moment in the history of computer science, transforming ethical considerations from an afterthought into a rigorous, multidisciplinary field. By bringing together AI researchers, legal scholars, philosophers, and economists, the conference established a roadmap for tackling bias, accountability, and the socio-economic impacts of automation.

Motivation: Why Now?

For decades, AI research focused primarily on performance—speed, accuracy, and scalability. However, as the authors Benjamin Kuipers and Nicholas Mattei note, the rapid integration of AI into the real world has created a "comparative advantage of automation" that threatens traditional labor structures and introduces systemic biases. The motivation behind AIES-18 was to create an intellectual "town square" where the technical "how" meets the ethical "why."

Methodology: Crossing the Disciplinary Divide

The conference was structured around four focal pillars:

1. AI & Jobs: Analyzing economic inequality and labor displacement.
2. AI & Law: Focusing on governance and responsibility in autonomous systems.
3. AI & Philosophy: Exploring value alignment and superhuman intelligence.
4. AI Technical Research: Developing algorithms for fairness and explainability.

The Moral Machine & Value Alignment

One of the most discussed methodologies was the Moral Machine Experiment presented by Iyad Rahwan (MIT). This experiment utilized crowd-sourcing to map human preferences in unavoidable accident scenarios for self-driving cars.

Figure 1: This experiment highlights the tension between descriptive ethics (what people do) and prescriptive ethics (what machines should do).

Key Discussions and Experimental Insights

The conference moved beyond theory into practical technical challenges. Several highlights included:

• Algorithmic Bias: Research from McGill University addressed how data-driven dialog systems often inherit and amplify hate speech, necessitating new detection mechanisms.
• Explainability (XAI): A session was dedicated to generating human-readable explanations from the "black box" of Deep Neural Networks.
• Social Governance: The ACLU emphasized that technologists must move beyond the lab and advise legislators to ensure "Liberty and Justice for All" in the era of ML-driven criminal justice.

Figure 2: The conference featured vigorous discussions across diverse panels, including representatives from IEEE and major tech labs like DeepMind.

Critical Analysis & Conclusion

AIES-18 was likened by Benjamin Kuipers to the 1973 IJCAI—a "proto-field" full of energy but lacking a unified framework. While the papers presented were fragmented across various disciplines, the conference successfully identified the "Inductive Bias" of our current systems: they are only as ethical as the data and the intentions of their creators.

Takeaway: The "Because we can!" attitude of developers must be replaced by a framework of moral responsibility. As AI continues to grow from a niche academic pursuit into a global industrial force, the insights from AIES-18 serve as the foundational ethics code for the next generation of intelligent systems.

Limitations: A recurring critique during the sessions was the "unrealistic abstraction" of current ethical scenarios (like the Trolley Problem). Future work must focus on more nuanced, real-world deployments where ethical trade-offs are not binary but multi-factorial.