[JamilD]

I’m very skeptical about the veracity and accuracy of facial recognition software to detect emotions.

I went to an affective computing conference in 2019 and was underwhelmed; models couldn’t distinguish between looking upwards (and raising your eyebrows) from exhibiting surprise. Emotions are complex and personal, and the phrase from the article “facial expression software — which nearly eliminates possible bias” seems absolutely ludicrous to me.

I take results like this with an absolutely massive grain of salt, and don’t expect them to be reproducible.

The danger is that they’re “catchy”, clickbait-y results, that are popular because people like to hypothesize about underlying psychological reasons why those bronze medalists might be happier. But let’s examine the core claim first, and not take facial recognition software as a ground truth for emotional state.

[rflrob]

In this case, however, there is evidently prior research using humans judging the faces that shows similar results. There definitely are all kinds of biases that could creep in from using facial recognition software, but a well designed study will attempt to quantify and control for those biases.

I definitely agree that the idea that software eliminates bias is laughable. Swaps it for another, hopefully less extreme set of biases doesn’t draw as many clicks, though.

[kapp_in_life]

>prior research using humans judging the faces that shows similar results

I always assumed that was generally what the models were trained from, are they not? Agree though that ai will just replicate the bias in the training data, which is why its important to have "bias free" data. Or at least as much as such a thing exists.

[IshKebab]

I think he meant humans judging Olympic medalists specifically.

[rank0]

Excellent point. People need to be more skeptical of novel technology in general. I mean, the big proponents of any new technology usually have a financial incentive to over sell its capabilities. I've noticed that with AI/ML especially people are willing to hand-wave away nuance and blindly believe that anything is possible. In reality AI is mostly iterative pattern matching and is far from perfect.

In this case however, I do believe that bronze medalists are happier than silver medalists provided that all the participants had a similar probability of victory.

[necovek]

> I do believe that bronze medalists are happier than silver medalists provided that all the participants had a similar probability of victory.

But that's never the case: betting companies are good at earning money off of that.

[johnsmith4739]

You are right in being skeptical.

First, behavioural responses vary from person to person and from culture to culture (Americans will smile in most social situations, Russians don't smile unless there is a laughter coming in).

Second, we are far from associating an internal subjective experience (joy) to an external objective behaviour (smile). It's actually one of the greatest challenges in psychology today. Also, how do you gauge my level of happiness compared to yours?

Third, the whole article is superficial and they don't hide it: first they measure who smiles more ... then they jump to the conclusion - all bronze winners perform the same cognition. Pure speculation.

Even if smiles would denote an emotional episode (joy) it would have to be right in the moment of the emotional episode - and it would be a micro expression (under one second).

Quote: "they studied medal stand photographs" -> posed circumstances.

[mcswell]

I was going to say the same thing: There's no way facial expression software (not quite the same as facial recognition software) is un-biased. What unbiased ground truth could you possibly train it on? Answer: I don't think there is anything, at least nothing that is valid across cultures, genders, races and so forth (and I'm skeptical that there is any good training data even for white college age males).