Hacker News Reader

Comment by smath | Hacker News Reader

There is a great blog post [0] that sohl dickstein had a couple years ago, about what happens when the metric being optimized is only a proxy for the thing you want to maximize — what happens eventually is “overfitting”. The original metric (which we cannot directly optimize for) actually starts to get worse.

I believe publishing papers and citations are those proxy metrics.

[0] https://sohl-dickstein.github.io/2022/11/06/strong-Goodhart....

View on HN

The article here correctly points out another thing though: society/academia rewarding a fundamentally wrong thing.

This isn't just "misaligned rewards", it's academia pretending to reward one thing when really it does something else entirely.

As the author observes, objective truth is no longer socially rewarded. Pretense is all that matters and accepted as "social truth".

I don't like objective truth as a concept, because it makes truth a matter of belief and easily turns into just another name for social truth. If you believe that something is objectively true but new evidence challenges that view, people will often ignore the evidence, because it goes against their established views.

Science, and research in general, should mostly be concerned with effective truth. The best objective-ish explanations so far, but something that can always be challenged with new evidence. The difference to objective/social concepts of truth is that you should always be willing to question the assumptions, beliefs, and reasoning ability you and people who agree with you have. If you discover something that goes against established beliefs, you should always start by assuming that you made a mistake somewhere. And if other people discover something like that, and it looks like they are seriously questioning the discovery, you should also take it seriously.

That's remarkable nonsense.

"Objective" truth is what you can derive from physical reality by logical arguments. The very opposite of "belief".

Social truth is what your social peer group wants you to believe.

"Effective" truth I've never heard about and your explanation merely describes a simplistic popular scientific method concept. Leading ideally to the objective truth I was talking about.

The internet describes "effective truth" (or "effectual truth," from the Italian la verità effetuale) as a philosophical concept introduced by Niccolò Machiavelli in his book "The Prince". It means judging people, leaders, and situations by their actual actions and real-world results rather than by their stated words, ideals, or moral promises.

Logical arguments can only be used in theoretical models. Whether the conclusions made within a model are also valid in reality is an empirical question that can ideally be tested with falsifiable predictions. But no matter how many falsification attempts you make, you cannot prove that your model is a correct description of the real world.

True, but you can get pretty good estimates that suffice for real life.

I'm not a big fan of gravity either, but it will happily kill me if I make a wrong move, whether I like it or not.

Yes, the proxy (here, papers & citations) becomes the goal, and so we end up in strange places :(

Reply on news.ycombinator.com