GAME THEORY OF MEMORY: RECALLING DISCOVERED POINTS, A MACHINE AGAINST A HUMAN
This note completes a series of three games between a human and a machine. In the first game, inventing a point, the human wins; in the second, reacting in time, the robot wins. This note proposes the third game: going back to a point that has already been discovered. The human forgets, so the information that its memory keeps about the point can only decrease with time, and its uncertainty about where the point was grows. A machine that keeps the point in its context recovers it with a probability that does not depend on the time elapsed. In the limit of an unlimited context, there is therefore a crossover lag beyond which the machine recalls the point better than the human, and an uncertainty horizon beyond which the human can no longer hold a conjecture about it. With a finite context, the machine's advantage holds only until the point is evicted. We state the model in plain words, give three hypotheses and three conjectures, compare the three games, and list open problems. Appendix A contains a preliminary formalization with proof sketches, based on the data processing inequality, Fano's inequality and an erasure model of forgetting; it has not been verified by a specialist, and the author invites researchers to complete it.
Authors
- Javier Gutiérrez-Maturana Sánchez (ORCID: https://orcid.org/0009-0002-0449-9825)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24426312
- Primary Topic
- Memory Processes and Influences
- Type
- preprint