Definition
A tentative representation is a problem representation held loosely — not committed to early, updated as new evidence appears, and dropped if a better representation emerges. In Hofstadter's GEB Topic 18 and Fluid Concepts and Creative Analogies (1995), tentative representations are a core cognitive capability and a missing ingredient in 1970s symbolic AI.
Why it matters
How it works
Maintain multiple candidate representations of the problem in parallel. Let evidence flow in and update each candidate's strength. Periodically reassess: is any candidate clearly winning? Is one clearly losing? Should a new candidate be generated? Drop losers, refine winners, generate new candidates when stuck. The discipline is not commitment-aversion — it is patient commitment, with a willingness to update when warranted.