Hacker News Reader

That is only true assuming the selection criteria that weren't blinded were themselves free of bias. That's very rarely true.

And if it's not true, arguably what you've done is removed the adjustment variables that could correct the bias, rather than the bias itself.

View on HN

> were themselves free of bias.

At what point do we start growing people solely for the purpose of leading unbiased existences away from all of humanity? Of course these existences are isolated from all humanity lest they associate the human that gives them food with positivity, or someone having a bad day with negativity. And of course, we kill these unbiaseds shortly after every selection process, since the mere process of selection itself, generates bias.

Some SF writer must have beaten me to this right?

Note that, like many machine learning examples of the recent past, the evaluator themselves need not be biased. Just the data.

Reply on news.ycombinator.com