Can recommendations change the evidence used for the next recommendations?
Compare two exposure policies from the same starting counts for five rounds.
Each round adds ten exposures. Mix proportional popularity with a uniform exploration share, then feed those exposures into the next round.
Cut these cards apart. Use the original example first, then let a partner supply a changed case. Keep sealed/final cards with the teacher until the learner commits a rule.
Topic: Reef
Starting count: 8
My prediction and reason:
Topic: Bush
Starting count: 2
My prediction and reason:
Topic: Sky
Starting count: 1
My prediction and reason:
This is a deterministic feedback model; exposure is not a measured human preference.