Society-in-the-Loop: Programming the Algorithmic Social Contract

How should oversight work when an algorithm affects people beyond its operator? Iyad Rahwan proposes Society-in-the-Loop, combining human feedback with an algorithmic social contract.^[1]

Contribution and argument

The framework asks stakeholders to negotiate value tradeoffs and the distribution of benefits and costs, then monitor whether the system respects the agreement. It identifies gaps in articulating values, quantifying externalities and verifying compliance.^[1]

Alignment relevance and limits

This places governance inside the feedback loop. Social Alignment includes whose interests matter; Preference Aggregation supplies some decision mechanisms, but selecting a legitimate mechanism remains a political problem. Training Socially Aligned Language Models in Simulated Human Society uses simulated feedback and does not implement this stakeholder-negotiation proposal.

The paper offers a conceptual agenda rather than a complete institutional design or deployed assurance method. Representation, enforcement, difficult-to-measure harms and evolving values remain open. Its social-contract analogy should not be mistaken for demonstrated public consent.

Historical context

Iyad Rahwan submitted Society-in-the-Loop: Programming the Algorithmic Social Contract to arXiv on July 23, 2017.

The proposal connects human oversight with stakeholder negotiation and monitoring. Society-in-the-Loop provides the concept entry; Social Alignment explains why affected people matter beyond the immediate operator.