[claar]

I started using SalaryFairy after seeing your link here -- nice site!

I understand the "collective intelligence" angle you're going for. However, the way SalaryFairy is implemented seems to "taint" the collective intelligence, at least in my case.

After predicting quite a few salaries, my score was -9 because apparently my predictions were different from the "crowd" predictions. This negative score is encouraging me to "guess" the crowd predictions, rather than simply make my best guess as to what the salary would be for the given prediction.

Since I noticed that my guesses are, on average, quite a bit lower than the "crowd" guesses, my new "strategy" is to guess the salary, then pad it $10-$50k to improve my score.

This seems like a bad thing when it comes to the "collective intelligence" goal of the site.

Just a thought -- hope it's helpful.

[cagriaksay]

It's indeed very helpful. Predictions should be independent for wisdom of the crowd to actually work. But we did introduce this dependency by revealing crowd predictions after yours as a tradeoff. We want to give feedback on how our users are doing, so that they're incentivized to continue making predictions.

Sometimes what you see as crowd prediction could be just a couple predictions. As new predictions come in, we adjust individual scores. We are planning to hide scores until a healthy number of predictions are made on that profile. That should help a bit on the dependency issue. We are open to suggestions as well. Thanks for the feedback.

[jghn]

It seems way off for my area (Boston). It's prediction for me is about 15K low and I know I'm fairly underpaid based on offers I've received from other companies (I like where I am and am fine w/ the low salaries). I found a similar pattern when I went to do predictions, the crowd's wisdom simply didn't seem to match up with the reality that I have seen.

And of course, as the GP stated, once I started getting dinged for guessing higher than the predicted score, I started lowering my predictions. All the scoring system managed to accomplish was to drive my predictions closer to the standard data, which seems like a bad way to motivate people to give honest answers.