Are you a good employee? Let’s check with the team

By Tyler Cowen

Working in ever larger teams, often enabled by technology, is a fundamental feature of contemporary life. And yet the broader social and economic implications of this trend are rarely discussed — especially when, as will become increasingly common, one of the collaborators is technology itself.

Consider business. For decades now, big businesses have been on the rise in the US, which means employment in large corporations that use a team approach is increasingly likely. One effect of this is that individual outputs are harder to measure. If a product does well, it is often not clear who should get the credit, because the inputs of so many people were involved in creating it.

It is difficult to recalibrate incentives to reflect this changing reality. Often companies respond by enforcing greater credentialism, trying to ensure that everyone is a worthwhile contributor. That could involve looking for an Ivy League education or a standout GitHub profile. Either way, companies are more likely to look for ex ante signals of quality and less likely to take chances on true outsiders, because if the outsider isn’t pulling their weight, it might not be evident for a long time.

The opposite scenario would be that of a chess player or a tennis star. They too work with their teams, which they are careful to thank when they win — but if they do win, the players take home a large share of the prize money. And in these fields there is very little credentialism. Magnus Carlsen was just a young kid from Norway who kept on winning games and rose to the top. He never needed a master’s degree in chess.

The real losers in the team system are those who do not have the temperament for all the schooling and credential-gathering. Those credentials of course include recommendations from well-known contacts, so networking and socializing have become increasingly important. This is a workable situation for most people but a frustrating arrangement for others.

Some recent evidence indicates this problem is especially serious in the world of science. The number of authors on scientific papers has been rising sharply, a trend I have observed in my own field of economics. It was once rare for the research paper of a fresh job-market candidate to be co-authored; now it is common. The work may be wonderful, but how can you tell how much any one author contributed? In the natural and biological sciences, one paper can have dozens of co-authors.

Again, credentialism will become more important, not less. In relative terms, someone from MIT listed on a multiple-authored paper is more attractive than someone from Iowa State University.

The obsolescence of current incentives may bite even stronger yet with the advent of AI and large language models. LLMs read and scan large amounts of data, and can be considered an extreme example of joint production. But they do not typically have access to most archival data, unless it has been put online. If the papers and letters of a famous scientist are held at a university, for example, the LLMs probably will not have digested them.

If an enterprising historian wrote a book based on those and other papers, it would count toward academic tenure. But say the historian instead put those papers into a form that could be read by the major AI services. Even if all issues of legal permission were handled deftly, that researcher would not receive much of a career reward. After all, his or her name is not on the LLM. And yet the LLM database, now including these additional papers, would be more valuable.

Also consider that the book might cover only a single substantive thesis, and as an academic volume it could cost $100 or more. In contrast, the newly enhanced LLM could respond to all sorts of questions about the data it has absorbed.

Current academic institutions — come to think of it, current societal institutions in general — under-reward people who improve the quality of LLMs, at least if they work outside of the major AI companies. This does not feel like a big problem at the moment, because people are not used to having quality LLMs. But moving forward, it may slow AI progress considerably. Scientists and researchers typically do not win Nobel Prizes for the creation of databases, even though that endeavor is extremely valuable now and will become even more so.

Adam Smith was amazingly prescient about the division of labor in a market economy. But one aspect of modernity that he did not anticipate was the tendency toward greater credentialism, and the growing difficulties of rewarding co-creators.