tag:blogger.com,1999:blog-7782629678785397724.post7786585763662051179..comments2023-09-29T05:45:18.939-07:00Comments on Scientific Pearls of Wisdom: Discrimination = Overgeneralization Max Wellinghttp://www.blogger.com/profile/06800595470095643387noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-7782629678785397724.post-79101172801829878312018-10-05T22:20:17.563-07:002018-10-05T22:20:17.563-07:00I think things like this are really interesting. I...I think things like this are really interesting. I absolutely love to find unique places like this. It really looks super creepy though!! <a href="https://www.credosystemz.com/machine-learning-training/" rel="nofollow"> Best Machine Learning institute in velachery </a> | <a href="https://www.credosystemz.com/machine-learning-training/" rel="nofollow">python machine learning course in velachery</a> | <a href="https://www.credosystemz.com/machine-learning-training/" rel="nofollow">Machine Learning course in chennai</a>Anonymoushttps://www.blogger.com/profile/13662377101256298223noreply@blogger.comtag:blogger.com,1999:blog-7782629678785397724.post-48524089155983285672013-07-25T13:27:17.942-07:002013-07-25T13:27:17.942-07:00It seems to me that whether or not you're over...It seems to me that whether or not you're overgeneralizing when you stereotype depends on what features you have available. If you are trying to hire someone based on their CV, you have a lot of relevant features besides race and gender, so if you only use the latter you are indeed underfitting. <br /><br />But if you are walking down a dark street trying to decide whether the person coming the other way poses a threat, you have only superficial features to go on. And if you conclude that they are "shady", the features that made you do so, e.g., visible tattoos, probably are correlated with the threat level.<br /><br />I think the problem with stereotyping is not that it necessarily overgeneralizes, because that's only sometimes the case. The problem is that even if the stereotype is correct, it is unfair to person being stereotyped. People with tattoos should be just as innocent until proven guilty as anyone else, regardless of the correlation between tattoos and criminal behavior.<br /><br />I once saw a talk by an engineer doing fraud detection for Apple. He said their machine learning algorithm wanted to flag all orders from a zip code in south central LA as suspicious. The problem is not that this is a bad rule (on the contrary, it's probably a great rule). The problem is that it's a deeply unfair rule.Shimonhttps://www.blogger.com/profile/01094897493062502808noreply@blogger.com