Source reconstruction · Everything here is proposed. An AI prepared this map from supplied documents, with page references on every statement and connection. The source author has not reviewed or endorsed it. Sources and limits →
Reason Commons
Source reconstruction · Proposed

Evolving Prosocial AI

From Paul Atkins and David Sloan Wilson’s Evolving Prosocial AI · reconstructed by AI

Published 26 October 2026
Release 2026-10-26.1

Open questions

5 published items with open questions, grouped by the kind of work they need. An item can be accepted and still have work left on it.

Check an assumption 1

Test whether capability design must cost a firm against competitors.The authors' answer is to change the institutions, so cooperation pays (pp. 36–37, 45); B Corps are cited as matching conventional firms financially (p. 43).Connection“Optimise products for engagement and growth, as competitors do.” conflicts with “Design products to build users' capabilities, even at some cost to engagement.”Where do needs pull two ways?

Other unfinished reasoning 4

Find one capability measure two assessors would agree on.The authors say the measurement frameworks needed to track capability impacts across scales require further development (p. 48).ConditionCapability impacts can be measured.What are we trying to achieve?Check the studies behind the erosion claims.Several cited findings are preliminary: Kosmyna et al. (2025) is under review and Fang et al. (2025) is a preprint (pp. 21–22). None has been checked here.Connection“AI is built to capture attention and to complete tasks for people rather than to build their capabilities.” contributes to “People's competence, autonomy and relatedness erode: critical thinking declines with heavy AI use, choices are delegated to recommendation systems and chatbot companionship displaces human contact.”What stands in the way?State one prediction across scales that could come out false.The authors say the examples show proof of concept at specific scales and that integration across all levels awaits rigorous testing (p. 37).Desired effectSuccess at each scale reinforces the others.What will we change?Find someone ready to take one step.No researcher, technologist, policymaker, funder, owner or date is assigned. Existing conditions are illustrative, not observations.ActionResearchers develop capability measures.Who does what?

These come from the commons' own record of unfinished reasoning, and from disputed items. No deadline, owner or priority is shown because none is recorded.