3 ms·

As an academic, I don't think this would stick as it makes it hard to refer to papers in a common way (if you're chatting with your team). Referring to somethi

As an academic, I don't think this would stick as it makes it hard to refer to papers in a common way (if you're chatting with your team). Referring to something like the "Lin transformers paper" or the "Vaswani attention paper" is easy to get people onto a common ground. If you randomise the authors, you'd need to fall back to full paper titles, and long paper titles aren't particularly evocative. Randomisation may be fine for citation (particularly numbered citation formats like[0], where the names are only in the bibliography), but even then, you are breaking the flow of thought. If someone references an article like "blah blah(mayford 2020)", then for those familiar with the field you can more easily continue reading without having to break your chain of thought to jump to the bibliography and figure out what paper that was. edit ---- a further thought... referring by TEAM might be alright, but falls down when you're multidisciplinary. e.g. you might refer to the "Deepmind" paper, or the "Facebook llm paper"...but my area of research is agritech so we work with animal scientists from a different institution. Do we refer to both teams? 'The google/meta paper'? Again, you're likely to just fall back to whatever is the first author or institution in your mind.

Maken 3y ago
What about referring to venue/publication, like the "NIPS attention paper"?
vivegi 3y ago
I agree. The "AKS paper" or "AKS primality test" or "AKS PRIMES in P" paper is a good example of the recall value of human names/initials. The initials are alphabetical order (of last names) and it provides enough context to do offline lookups.
8n4vidtmkvmk 3y ago
> Again, you're likely to just fall back to whatever is the first author or institution in your mind. And that's exactly the problem we're trying to solve :-) If only the first author is in your mind, then the rest of them aren't getting enough credit. Next time you're reading a new paper and it's written by the 4th author, are you going to recognize their name and go "Oh, this was written by that author that wrote that other great paper, I should read it!"? Or will you not have a clue who they are because only the first author stuck out? Disrupting the thought process so that you either need to refer to all of them or their team or the paper by name is the point. If it's written by 2 teams, it's the same problem. You would have to randomize Google/Meta and Meta/Google. Although..well maybe not if it's short enough that it's always referred to in full. With regards to paper names being too long, we have that problem at my office sometimes too. I just give them cute shorter code names so they're easier to refer to.

Maken 3y ago

What about referring to venue/publication, like the "NIPS attention paper"?

vivegi 3y ago

I agree. The "AKS paper" or "AKS primality test" or "AKS PRIMES in P" paper is a good example of the recall value of human names/initials. The initials are alphabetical order (of last names) and it provides enough context to do offline lookups.

8n4vidtmkvmk 3y ago

> Again, you're likely to just fall back to whatever is the first author or institution in your mind. And that's exactly the problem we're trying to solve :-) If only the first author is in your mind, then the rest of them aren't getting enough credit. Next time you're reading a new paper and it's written by the 4th author, are you going to recognize their name and go "Oh, this was written by that author that wrote that other great paper, I should read it!"? Or will you not have a clue who they are because only the first author stuck out? Disrupting the thought process so that you either need to refer to all of them or their team or the paper by name is the point. If it's written by 2 teams, it's the same problem. You would have to randomize Google/Meta and Meta/Google. Although..well maybe not if it's short enough that it's always referred to in full. With regards to paper names being too long, we have that problem at my office sometimes too. I just give them cute shorter code names so they're easier to refer to.