It is often true that we shouldn't judge a book by its cover, and that we shouldn't judge people just by first impressions. The problem is the same people who say that are also the same people who say that first impressions matter. But, of course, first impressions don't matter to them, just to other people (if you didn't get the snark from that sentence, carry on). I am often chided by people I know for judging others too quickly (despite the idea that first impressions matter, y'know) and, to some people, I am the most judgmental person they know.
I would like to offer a counter-perspective. First, let me both offend and defend those who say we shouldn't judge people by their first impressions but simultaneously suggest that first impressions matter. To defend, you are not completely wrong. To offend, you just haven't thought it through enough. I think I can put forward a way to think about this that makes sense and does not make you (or me) a hypocrite. In other words, I'm doing your thinking for you. If that sentence sounded super obnoxious, please note that it was happily laced with snark and thus, meant to be super obnoxious.
Just to lay my cards on the table, part of my opinion on a person, or a matter, is shaped by my first impressions or observations of them. So, yes, by the definition of most, I do 'judge' people by their covers. If someone says something stupid, my initial opinion is formed, based on that observation. However, the key difference between my definition of 'judge' and the common definition of 'judge' is that all I'm doing is making a hypothesis. This, I think is using the Scientific Method to evaluate people.
The Scientific Method to People is based on three foundations. The first is the Scientific Method (duh). The second is the concept of Bayesian Priors. The third is the aggregate baseline. I will go through each in step.
Step 1 - The Scientific Method
This is essentially treating each individual as a potential experiment. I first observe them when I first meet them. Based on those observations, I make a hypothesis about them. A hypothesis is essentially something to test; it is very different from a conclusion which is ex-post. From that hypothesis, if I have more interactions with these individuals, I therefore update my hypothesis based on the new data and information that I get from 'experiment.' At some point, you hit a diminishing marginal returns of new knowledge segment when meeting a given individual. (For example, you are much less likely to find out something surprising about someone you've known closely for 10 years than you are someone who you've known for just 2 weeks). When the curve essentially becomes relatively flat, then you can make a conclusion.
Look, for everyone we know, we do make "judgments." But it's important to know which ones are hypotheses and which ones are conclusions. It applies to everyone - friends, romantic partners, FWBs, colleagues, affiliates etc. The more you meet them, the more 'experiments' you conduct on them. For some, you'd rather stop after the first 'experiment'. For others, you want to undertake more 'experiments'. The Scientific Method is all about experimenting and figuring out what the data tells you. What the data tells you should be your conclusion. But, how do you know how many experiments to run? What is the requisite number? The way, I think, to answer that question is through the Bayesian Priors.
Step 2 - The Bayesian Priors
To evaluate the probability of a hypothesis (in other words, whether your hypothesis works), the Bayesian probabilist specifies some prior probability, which is then updated in the light of new, relevant data. Essentially, one updates the probability estimate for a hypothesis as new information gets received. What this says, for the purposes of the Scientific Method to People, is that it's totally fine to make hypotheses about people (A given person is likely kind, stupid, stingy, immature, brave, optimistic etc etc) as long as you are willing to allow your hypotheses to be changed as data from experiment leads you in different paths. You cannot be too hooked onto your hypotheses such that you are unwilling to change them. Precept 4 of Prospero's Precepts says that one should never fall in love with one's hypothesis.
So, it's okay to make evaluations of people based on your observations. Everybody does that (for the record, judging doesn't always imply negative judgment. Judging that someone is smart is also judging). But you should always be ready to change them. Not everybody does that. So, update your priors as you go along.
So, we have established that a) Make hypotheses, run experiments and then make conclusions and b) As you run experiments, be ready to update your priors and thus, your hypotheses, based on new data. Now, how do we form hypotheses? What would be an objective way of forming a hypothesis? For this, we need the aggregate baseline.
Step 3 - The Aggregate Baseline
Consider this quote from Sherlock Holmes in "The Sign of the Four." He says, "...While the individual man is an insoluble puzzle, in the aggregate he becomes a mathematical certainty. You can, for example, never foretell what anyone man will do, but you can say with precision what an average number will be up to. Individuals vary, but percentages remain constant. So says the statistician." Sherlock Holmes is essentially describing the Law of Large Numbers, a statistical concept which posits that if the sample size is large enough (i.e. you've performed the experiment often enough), the mean of the experiment will be the expected value.
Let's take an example. Consider the toss of a fair coin. The probability of heads or tails is 0.5. Now, if you toss a given coin, you don't know whether it'll be heads or tails. You just know that it's a 50-50 chance for either. So, the individual toss is unpredictable. If you toss it 10 times in a row, you could still get 10 Heads or 10 Tails, based solely on complete randomness. Consider the following sequences of tosses:-
1. H H H H H H H H H H
2. T T T T T T T T T T
3. H T H T H T H T H T
4. H H H H H T T T T T
5. H H T H T T T H H T
6. T T H T T H H H T H
With a fair coin, the probability of each ot these 6 sequences is exactly the same - 1/2 to the power of 10. While 5. and 6. may seem like the most 'random', 1, 2, 3 and 4 are equally likely as well (please see the end of this blog post for a note about intelligent design). The Law of Large Numbers, and Sherlock Holmes, says that if you do things at a large enough level, you should get your expected probability which, in this case, is 1/2. So, for a large enough sample, you should expect Heads to show up 50% of the time and Tails to show up 50% of the time, even if this is not what it ends up being.
AND SO, with people, it's the same. I find it acceptable to use stereotypes (the aggregate baseline), even if they are politically incorrect stereotypes, as a way of forming my initial Bayesian priors of people. But that in no way means that I conclude that about a person, and thus, do not change my mind. For example, if I see someone who spends a lot of time in the library, I'm likely to think that she (assuming she's a girl) is the love of my life a studious person. And that would fit the stereotype aggregate. And that's how I form my hypothesis. As I go on dates with her interact with her more, I update my priors to the 'individual' level. Maybe she's not the love of my life studious. Maybe she only spends a lot of time in the library for a reason other than something studious. And that's fine. That's how I update my priors based on new data, and that's how I form a conclusion. I just must not fall in love with my hypothesis.
So, how many experiments to run? I actually would argue that you can never really make a conclusion and that everybody is just an ongoing 'experiment'. I think that's fair - you may not have to experiment more, you can prioritize some 'experiments' over others. But that still leads back to the same problem - how do you choose who you want to get to know better and who you don't? Again, it's the aggregate. Most people tend to fit into particular types of stereotypes, and you can decide, for yourself, if you want to hang out with that type of person - party animal, nerd, jock, geek - while being, firstly, fully aware that a person may not be the stereotype you think, secondly, fully aware that a person may be more than one stereotype, thirdly, fully aware that you may be completely wrong and, fourthly and most importantly, you do not treat someone with less respect and less dignity just because they happen to fall into a certain type of stereotype, particularly arbitrary stereotypes. 'Arbitrary' stereotypes here mean stereotypes that people fall into by no means of their choices - race, ethnicity, sexual orientation, gender. It is acceptable and reasonable to not want to experiment further with someone who knowingly smokes (a smoker stereotype). It is completley unacceptable and unreasonable to not experiment further with someone just because they're Jewish or gay or male or whatever arbitrary attribute.
Therefore, feel free to stereotype people but keep in mind the four caveats I had earlier. You can use stereotypes to form your hypotheses but you should never use stereotypes to form conclusions. You can decide for yourself if you want to continue the experiment or conclude there and then, but I'd strongly recommend keeping all hypotheses open and allowing your Bayesian priors to be updated accordingly. All we can do is to just use the aggregate baseline as our priors, updating our priors as we go along, choosing the stereotypes that we want to get to know better. For example, I'd rather not hang out with someone who racially profiled people at our recent elections because I think that person is a racist. But, I'm open to the possibility that that person might not be racist (though I'm probably right). Yes, caveat - you can use the "I'm probably right" notion if you believe yourself to be a good judge of character. So, how do you choose how many experiments to run? Choose the stereotypes or aggregate baselines that you like and experiment more with those. You don't have to experiment more with the stereotypse you dislike, but keep in mind the caveats earlier.
Lastly, you know how in movies or books, the potential romantic partner goes, "I really like you. You're unlike everyone I ever met," or "I really like you; I can't figure you out and that excites me." It's really cliche but it's representative, actually, of everything I just typed. It's essentially someone updating their Bayesian priors about their interest and finding out that this person doesn't fit into the stereotype that the Bayesian thinker first conceived. And when you can't put that person into a given bucket, the person becomes intriguing. Like Precept 7 in Prospero's Precepts, "The thing that doesn't fit is the thing that is most interesting."
----
Quick note on countering the argument on intelligent design. Consider:-
1. H H H H H H H H H H
2. T T T T T T T T T T
3. H T H T H T H T H T
4. H H H H H T T T T T
5. H H T H T T T H H T
6. T T H T T H H H T H
If we flipped a coin a million times, getting a million 'H's is also possible, if supremely unlikely. But, you can still get it through sheer randomness. Just because it looks pre-ordained doesn't mean it is. It is equally as likley as any other sequence. If the coin had fallen another way, the outcome would be different, but the process is the same. This is why the intelligent design argument fails, in my opinion. Just because things are the way they are doesn't mean they were intelligently designed. The probability for things being the way they are are just as likely as the probability for things being another particular way. Remember, you cannot conclude what the individual will do, even if you can mathematically estimate what the aggregate will do.
No comments:
Post a Comment