COMSCI/ECON 206 · Behavioral demonstration
Stay or Switch?
A behavioral demonstration of AI memory portability and switching decisions
Evidence Boundary
This is an exploratory classroom demonstration, not a representative or causal study of AI-assistant users. Its numerical benefits and hurdles are normalized parameters, not estimates for real platforms. A decision that differs from the benchmark is not automatically irrational or biased.
Your scenario
Imagine a possible switch
Imagine that you have used Assistant A for two years. It has learned your writing preferences, recurring projects, and interaction habits.
Assistant B offers a normalized expected benefit of compared with staying on A. This may represent expected quality, usefulness, or fit; it is not a dollar estimate or a real-product claim.
Your memory-portability condition is .
Make your own decision first. A simplified benchmark will be revealed after your initial decision and reflection are recorded.
Step 1
What would you do?
Step 2
Self-reflection
Optional, maximum 300 characters. Please do not enter your name, email, account details, health information, or other sensitive information.
Simplified Benchmark
A comparison point, not a recommendation
- Normalized switching benefit
- Modeled switching hurdle
- Net benchmark advantage
- Benchmark prediction
Rule: Switch if g > r; Stay if g < r.
This is a simplified benchmark, not a recommendation about what you should do. It intentionally ignores trust, privacy preferences, habit, uncertainty, attachment to existing personalization, and other considerations not captured by these two normalized parameters.
Step 3
Your final decision
Optional, maximum 300 characters. This text remains only in this page's memory and is never included in peer summaries.
Peer comparison
Compare with Earlier Anonymous Plays in This Browser Session
Because this free Static Space has no shared backend, peer summaries include only earlier plays from the current browser session. They do not combine responses across different devices or browsers.
How this demo works
The demo samples uniformly from 12 fixed scenarios: Full, Partial, and None portability crossed with normalized benefits of 0.15, 0.30, 0.45, and 0.60. The corresponding hurdles are 0.20, 0.35, and 0.50, with no equality cases.
The initial choice and four 1–7 reflection ratings are recorded before the benchmark is shown. The benchmark predicts Switch if g > r and Stay if g < r. A final decision can then be submitted once.
Because this free Static Space has no shared backend, peer summaries include only earlier plays from the current browser session. They do not combine responses across different devices or browsers. No server-side response storage is used.
Connection to the Research Model
The Bayesian model asks whether platforms strategically supply portability. The auction extension models portability as lowering switching friction. This behavioral artifact asks a distinct question: when modeled technical friction changes, does a hypothetical user choose to switch?
The layers are linked but distinct. This demonstration does not validate the Bayesian equilibrium model, and deployment alone is not behavioral evidence.
Full evidence boundary and privacy policy
This Space is a classroom behavioral demonstration linked to a computational-microeconomics proposal. Responses are exploratory plays, not a representative sample of AI-assistant users, and the demo cannot establish population-level effects or causal relationships.
Numerical switching benefits and hurdles are normalized experimental parameters, not empirical estimates of real AI platforms. A Stay/Switch decision that differs from the benchmark is not automatically evidence of irrationality or status-quo bias; the benchmark omits many real preferences and risks.
This free Static Space has no shared behavioral-data backend. Peer summaries refer only to earlier plays in the current browser session. The app does not collect names, email addresses, account histories, IP addresses, device fingerprints, or sensitive personal information.