Hacker News Reader

But even if they are accurate and sensitive, how can we avoid the bias of humans? The one link I see repeated on Reddit (drugs) is a paper that iirc showed over 50% were human alerts

https://health.ucdavis.edu/welcome/features/2010-2011/02/201...

View on HN

I think the primary issue with drug sniffing dogs is that the handlers aren’t inclined to care about the veracity of the result and would probably prefer the dog to trigger a positive result because it gives them a free pass on probable cause. The dog is likely rewarded regardless as well which is why you get high profile cases of dogs that alert 90+% of the time. They were literally trained to do it.

The same thing can’t necessarily be said for cancer detection. There isn’t a strong motivation to skew the results by the handlers and presumably any positive response would entail further diagnostic testing. I suppose cynically you could apply the same logic so that medical providers can justify the use of more expensive diagnostic testing but I’d like to think it’s not quite that bad.

My aunt has a seeing eye dog that has led her in to traffic 0 times. My dogs reliably find game birds and ignore non-game birds in the same cover. The trick is to not put them into situation where pleasing their handler is at odds with the job they are tasked with.

Reply on news.ycombinator.com