Artificially Intelligent

Any mimicry distinguishable from the original is insufficiently advanced.

How well can people distinguish tastes?

| 601 words

Recently, my friend claimed that he could tell the difference between two brands of sparkling water. I expressed skepticism. After testing, I’m now moderately confident that he can.

The reason I was skeptical seems to be downstream of a tendency I have to assign less weight than other people on things like memory, sensory experiences, etc. I think memory is unreliable mostly because of the psychological literature on the inaccuracy of eyewitness testimony, which I have a passing sense of from reading psychology textbooks and a small amount of original sources a few years ago. Sensory experience unreliability is downstream of the inability of wine experts to reliably rate types of wine, the most extreme example being an inability to distinguish dyed white wine and red wine. Note that it’s unclear how much expertise these people had. Other observations that contribute is the literature relating the shape of the container of food to its taste, e.g. whiskey poured from a square bottle tastes different to people than whiskey from a round bottle. People also seem to be unable to tell the difference between Coke and Pepsi, even if they’re fans of one of the beverages above the others.

Thus the general belief: people are often very overconfident in their ability to distinguish the taste of things when they haven’t controlled for confounding factors like beverage container, expecation, etc. Thinking about this more, the eyewitness testimony thing doesn’t seem to have that much bearing on the ability to tell the taste between things, so maybe I believed some fake correlation there. It also seems plausible that none of these experiements replicate, although my own experiences seem to at least roughly corroborate.

My friend pointed out that this was inconsistent with my belief that I can tell the difference between Oreos and off-brand cookies, different types of chocolate, etc.

How should I update? I can think of a bunch of possibilities.

  1. Wine, in particular, is difficult to tell the difference between.
  2. Telling the difference between things when you’re trying is not that hard, but telling when things are the same is hard.
    • This is compounded by expectation. If you told the wine people that there was a chance they were the same wine, many of them would have gotten that correct.
  3. People are pretty good at telling the difference between different tastes. The results about people being unable to distinguish between pepsi and Coke are wrong, or pepsi and Coke are uniquely similar.
  4. Sparkling water has many dimensions of variation. It’s closer to the difference between wine and vodka than two types of wine
  5. I don’t really know what’s going on with taste and my abstractions are broken.
  6. Water has huge variation in taste because of the presence or abscence of trace minerals in a very neutral medium.
  7. Differences in carbonation/acidity are easy to detect.
  8. Coke and Pepsi are more similar than two brands of sparkling water.

I think my main mistake was something like extrapolating from “people can’t assign consistent scores to things” too strongly to “therefore people can’t tell things are different.” I think I also maybe underestimated the effect of tribal affiliation with Coke or Pepsi and assumed that people preferred one over the other because of taste. The more correct perspective seems to be “things like beverage container, expectation, affiliation, etc. can produce relatively large subjective differences in tastes between various things, but many subjective differences are not driven by these factors.” I think I also have a poor ability to taste things, so perhaps I am generalizing too much from my own experiences.