187 - Hiring Paradox

Do you judge a researcher's skills by how well they dance?

It sounds like a silly question, but that is precisely the sort of process that recruiters, investors, politicians, grant proposal analysts, and countless others make every day. Why this becomes a severe problem is also made clearer when you consider population dynamics over time, particularly within systems of competition.

Let us continue the metaphor and say that there is a population of 50 researchers all competing for a role, and they are being judged on their fitness for this role by their dancing skills. This has both immediate consequences and further consequences over time in the near and long term.

The immediate consequence is that those researchers whose skills are most purely and heavily focused on research will virtually never be judged as most fit for the job due to the fundamental misalignment of criteria.

The short-term consequences are that the poor decision-making structure based on proxies with weak or no correlation, let alone any causation, causes subsequent poor hiring decisions.

The long-term consequences come into play when this poor process becomes an industry norm, and many who aim for those research roles invest their efforts into becoming better dancers, rather than becoming better researchers. At this stage, the misalignment has become an entrenched and degenerative feedback loop.

Of course, every interview of a potential new employee and every pitch to a potential investor are processes of judging a candidate by their "song and dance" skills, and both processes are deep into that degenerative feedback loop. The HR domain has understood this problem for many years, and come up with countless proposals to treat it (absent any cures), but the investment space still seems to be only marginally and tangentially aware of the problem.

Just as a potential employee's resume-building skills have virtually nothing to do with skills required for most jobs today, graphic design skills and personal network connections to the wealthy have virtually nothing to do with the potential of most startups. The skills of "Song and Dance" to present oneself are good for the "Showmen" of Barnum and Bailey, but wholly inappropriate for serious scientific or financial decisions.

This is both a problem of immediate hyper-complexity and one with dynamics that change greatly (in this case negatively) over time. Neither humans nor trivial AI systems like LLMs and agent-based systems can handle hyper-complexity in any meaningful, sustainable, or effective sense, but there is a technology that can.

For those who prefer the song and dance, remember that eventually the Tiger decides to go solo.