Sometimes, when rating a film on IMDb, I give either 1 or 10 stars—not my honest rating—to maximally steer the average toward what I think it should be. Has anyone explored the dynamics of what happens if everyone always votes maximally or minimally to steer the average toward their desired value? Does this behavior have a name? I couldn’t find anything in a quick search.

I’ve done some thinking myself, and Monte Carlo^[1] simulations show a low average deviation (around 0.6/10 stars) between the steering equilibrium and the honest average when the population is split into a few distinct groups with different preferences, and then a stochastic generator spits out a new voter with probabilities ~ sizes of groups. It’s relatively straightforward to calculate where the vote would settle mathematically in this case.

If the groups appear in some order instead of being evenly mixed, this strategy favors the last group extremely heavily.

But even in an evenly mixed population, 0.6 isn’t 0, and I wonder if the golden rule would imply I should never do this for stable ratings (e.g., films). For dynamic things like restaurants, though, steering seems reasonable since their quality changes over time, and faster convergence to the “true” value might help everyone.

Would love to hear your thoughts on this.

1. ^{^}

   GPT-o1 generated code without unit testing or reviewing the code