So I went for [my college's] admit reception, which is essentially the most important hard-sell opportunity for the college to fill up the admit-matriculate yield (beyond so-called interest-selection bias]. Ah, the distorted incentive structures of US News rankings.
After the admit reception, I followed the alums to a pub where one of the topics of discussion was their role as alumni interviewers, and the interview component of the admissions process.
There is a cognitive bias among this particular group on the competitiveness of admissions – after all each of them got in. Having been through the admissions gauntlet a few times myself and with my friends, I was somewhat puzzled when one said that the interview process (which includes a written report on the candidate and a score) was very marginal in the overall selection process.
Wrong.
When the expected payoff is extremely high [signalling and other expected utility from admissions to your first choice], and the payoff only functions at a particular threshold (it doesn’t matter if you nearly got into your first choice, only whether you got in or not], then marginal factors matter. When competition quality is so high that differences between two candidates are often cosmetic, and the subjective and random components of what is supposed to be an objective valuation/selection process are systemic, then marginal factors matter.
Knowing that the interview component is, for all practical purposes, significant to a candidate’s chances, the burden of responsibility on the alumni interviewers to perform it in a fair and objective manner is so much greater. But to me, it seems like we are simply including more opinions, more random sampling, more noise:signal, more variation in implementation – more subjectivity and randomness, not more objectivity.
Let’s say that 2 candidates, ceteris paribus, are interviewed by 2 alums. Alum 1 is having a bad day, OR has just interviewed a high quality candidate, OR has a lower discount/inflation rate. Ceteris paribus, candidate 1 is more likely to have a less positive interview report and score. OR, noise is likely to be more important than signal: Is candidate 1 more attractive? Is candidate 1 from a similar socioeconomic/cultural background than alum 1? Does alum 1 have a greater google-presence than alum 2, HENCE is candidate 1 more prepared? Is candidate 1 better at interviewing skills, or has more experience? All of these are tangential to the primary traits being investigated, and raise the noise to signal ratio.
We have to recognize that there are problems with the present approach. There has to be a better way.
