Interactive notebook
Creator Incentives Lab
An immediate gain can change what people create next: adjust the platform’s choices, inspect the assumptions, and compare the trajectories.
Runs in your browser: nothing is saved and no account is needed. Optional AI features and Paper Q&A show their own notice and send nothing until you submit.
A simplified teaching model, not a reproduction of the paper. One price-taking creator, chosen functional forms, and dimensionless units; no strategic competition, empirical calibration, or welfare guarantee is simulated.
02Observe what changes
Effort today. Quality tomorrow.
Inspect the numerical output
| Period | Quality | Audience | Net profit |
|---|
03Ask a sharper question
An experiment guide that shows its work
A rule-based guide picks a bounded experiment, runs the same numerical program, and compares the results. It is not an autonomous AI agent.
Controls changed. This comparison keeps the parameters from its last run; rerun to use the new settings.
Ask an AI experiment planner
The planner can choose up to two comparisons, observe program-calculated results, and revise its next step. It cannot change the simulator, run arbitrary code, or use the internet.
The complete model and its limits
Retained traffic r = 1 - d(1-z). The creator chooses nonnegative effort to maximize u(e) = (r+b)e - ce^2/2, giving the exact best response e* = (r+b)/c. This is an optimization for one price-taking creator, not a Nash equilibrium calculation.
Initial quality q[0]=1. For periods 1 through 40, q[t]=0.8 q[t-1]+0.2 e*, audience a[t]=0.2+0.8 q[t]/(1+q[t]), and net platform profit p[t]=(1+0.7 d(1-z))a[t]-b e*. Total discounted profit is sum(t=1..40) beta^(t-1) p[t]. The numbers 0.2, 0.7, 0.8 and the forty-period horizon are illustrative choices, not estimates from data.
Citation returns traffic but reduces the assumed immediate AI-retention revenue multiplier. Compensation increases effort but is also a platform expense. Audience saturation, quadratic effort cost, and slow quality adjustment are assumptions you can question. No competition, heterogeneous creators, endogenous entry, or changing preferences are represented.
Read the actual AI Overviews paper for its game-theoretic model and findings. This toy model can suggest questions, but cannot verify its theorems or tell a real platform which policy to use.
Interaction inspiration: Distill's interactive explanation of momentum; no text or visualization code is copied.
Prepare a question for the AI Overviews paper assistant You review and send it yourself.