> If the answer is that contributions to scientific papers can be very fuzzy, e.g. it is hard to determine which contribution is more important than others, then I agree alphabetical order makes sense. But is it really so in most of the papers?

I'm probably biased from being in industry, but even when people contribute unequal code to a module or writing in a report, we just list the authors alphabetically because it's obnoxious to do anything with the list otherwise and it stops a huge amount of politicking when the first three out of ten authors contribute equal amounts. At the end of the day, stable collaborations will outpace individual contributors -- so we should focus on what allows us to openly collaborate and tears down barriers to doing that.

But from my point of view, we have a lot of the same incentives between industry and academia:

1. If you want to get promoted, you need projects associated with you and to get noticed.

2. There needs to be some method to decide who gets what amount of bonuses.

3. Funding allocation is tied to past successes.

...and probably some more I'm forgetting.

So I think my experience isn't totally irrelevant -- you get enough signal out of the distribution of the person's name that the loss of whatever from not getting "first" information is eclipsed by the gains from less politics (and if you have a lot of documents, easier parsing).

Again, perhaps it's the industry talking, but I'd rather hire an author that contributed the third most to a large number of papers than someone who contributed first most to a few. One of them seems to be a "force multiplier" in that they're keeping several other ICs moving (and contributing on all those efforts), while one guy just seems a specialist in a particular work.

I don't think removing the "first" signal harms my ability to find those "force multipliers", because they'll still have the same markers (as they weren't showing up first under the other system anyway).

So perhaps I simply have different uses for the rankings than people who feel differently?

I would suggest you consider the emergent properties of these systems of attribution, rather than just the primary affect. Could there be a reason why theoretical fields have small numbers of authors given alphabetically, and experimental fields sometimes have ten or more authors, given in order of "contribution"?

Isn't GitHub just ranking by number of commits? That's not a very good measure of contribution. After all, I don't think tensorflow-gardener is the primary contributor of Tensorflow.