The ethics of AI ethics board

Google dramatically dissolved AI ethics board just one week after forming it. Many ask who should be included in AI ethics board. Ethicists? Politicians? Who should seat in the board is an important issue, but what principles the board should use is a more important and fundamental issue. I propose the following principles for AI ethics board.

The framework is in line with the broadly Kantian/Habermasian procedural and discursive approach to justice and democracy. In particular, I am indebted to Daniels and Sabin's procedural approach to fair allocations of limited resources.

First, determining whether or not X (new product, policy, etc) is ethical or not should not be a matter of power or voting. Instead, any normative judgment should be a matter of reasonableness/legitimacy.

Second, X is permissible when all of the following five conditions are met:

  1. Utility Condition: X satisfies the Utility Condition if it is rational to believe (with the best theory and evidence) that such a change maximizes (or least enhances) the overall value of involved stakeholders.

  2. Generalization Condition (No-cheating principle): Company acts unethically unless it is doing something that it could rationally will every company to do, in similar circumstances, if it can.” (from Parfit)

  3. Publicity Condition: Decisions and rationales behind the decisions must be publicly accessible to all stakeholders. The Publicity condition is useful to ensure transparency of decision-making procedures, which allows involved parties to examine whether the decision-making processes are coherent, sound and grounded by evidence. Having the rationales publicly accessible can also make decision-makers to clarify and double-check their rationales and relate them to the people. Such transparency would ensure that decisions are open to scrutiny and debate by the people, which, in turn, can contribute to the quality of public deliberation and facilitate social learning. Decision-making processes that meet the Publicity condition demonstrate that the decision-makers are principled and responsive to the people, in particular to those who are affected by their decisions, thereby providing legitimacy to the decision-makers. If business secret is involved, a trusted third party can represent the public.

  4. Revision and Appeals Condition: There must be mechanisms for challenge and dispute resolution regarding decisions, and, more broadly, opportunities for revision and improvement of policies in light of new evidence or arguments.

  5. Regulation Condition: There is regulation of the process to ensure that conditions 1--4 are met. Ai ethics board that has no power is not a board.