A study on human behavior has identified four basic personality types
uc3m.es
uc3m.es
I don't know if this is true of the history of philosophy (or if it's a meaningful indictment of these philosophers), but I can't help but be reminded of it when I see strenuous efforts being made to bucket various phenomena, and human ones at that. What are we going to do with these categories? Not to put too fine a point on it, these things are complex. Any differentiating scheme simple enough will be insufficient to be the basis of any important decision, and any sophisticated enough won't be neat enough for us to be talking about it like this.
I say this as someone who spent more years than I'd care to admit obsessed with MBTI. Wow, look, there are all these different people with their strengths and weaknesses. I even get a cool framework—a stack—to aid me in psychoanalysis. And then of course you have Socionics, which shamelessly tries, using just sixteen categories, to produce a catalog of human interaction. I am not saying it isn't fun.
Actually, MBTI did help me in one way: by showing me how diverse humans are. In my most advanced, intelligent state, I'm still not going to get along with or be compelling to all of them. That was an important idea for me growing up, and still is.
But beyond that, and running the risk of sounding unsophisticated: who cares?
It felt like this lady was trying to shoe horn me into some category, telling me I am some 4 letter acronym and here are my strengths and weaknesses. I just kept arguing with her pointing out events and situations in my life that refuted everything she was trying to say I was.
It seems like an overly complicated horoscope thing where people love to read descriptions of themselves and others to make sense of the world, even though by looking at someone's face for a few seconds you can make way more judgements about the character of someone.
Complete baloney.
One of the big problems with bucketing is that personality is almost certainly continuous and probably gaussian as opposed to bimodal.
That means in nature, where is some normal distribution, there is often several normal distributions that overlay each other making it look like one continuous function with a noise rather than several continuous functions with different parameters. These studies are trying to uncover just that.
Years ago I was working with two guys - one had become a close friend and one was a guy that I didn't like at all, he totally rubbed me the wrong way. BUT the two of them were good friends, and so my friend kept asking me why I wasn't friends with this other guy since we both liked the same things, looked at the world similarly, etc.
So eventually I somehow discovered this book called "I Wish I'd Said That"[1], which categorizes and analyzes people according to their communication styles.
Now I had studied a bunch of other typologies (MBTI, Enneagram, etc), but all of the sudden this one made it clear that I didn't like my coworker because of how he communicated. I had leapt to conclusions about who he was based on his communication style, and realized I was reacting to the way he spoke, not him as a person. The book gave me a very clear way to reframe how I interpreted his communications and how to engage with him so I could get past the parts of his communication style that didn't match mine and actually get to know him as a person.
Long story short, we are now close friends - 100% due to reading and applying the categories in this book. Your mileage may vary, but I'm pretty sure it's hard to NOT learn something useful from it. Highly recommended, especially if you tend to look at the world through an analytic lens.
[1] https://www.amazon.com/Wish-Id-Said-That-Trouble/dp/04715555...
That seems to be an argument against anyone but experts ever discussing anything.
Simple models are useful, particularly as they can be understood by the less expert of us.
Categorization - to provide concrete value - requires feedback into the categorization scheme from real world data. There are various indicators which might point out that the categorization is useless or totally incorrect and then modify or abandon the categorization scheme.
For instance, in physics it is very practical to categorize collisions into elastic and inelastic collisions (the former conserves energy while the latter conserves momentum). But that's not all of physics, and does not even make sense all of the time.
And then we have categorizations like Meyer Brigs which are just nonsense.
The problem is, as you said, that people often have this strange illusion that just inventing a category somehow is somehow value adding.
Lucky that I work for a company that's data focused for arguments. And if I didn't, lucky enough to be financially stable to leave a company that's overly political.
Granted, that depends heavily on your features and your data set. It's just as easy to end up with one big undifferentiated blob, even after running PCA.
"It's not unlikely that you can say that your data can be partitioned into N distinct clusters."
Of course not...that's tautological. You can say it because that's what it means. DBSCAN isn't some wizardry that peers into the nature of the universe...its just a clump finder. Whether those clumps have any connection to anything interesting is a coincidence. You could roll a six sided die six times and run dbscan on the results and be like "holy shit! I've discovered how to partition the natural numbers 1 2 3 4 5 and 6!!"
Correlation does not imply causation, but you (generally) can't have causation without correlation. Finding natural patterns of associations in data is our only reasonable starting point for finding patterns of causation.
Also, developing reliable and well-researched ontologies can help other researchers when building models, making sense of other data, etc.
Feeding arbitrary sequences of samples into dbscan and deciding that because dbscan produces output, there is causation (or, that there is an underlying phenomena that can be captured in some type of model), is ridiculous. And there are tons and tons and tons of natural phenomena that will be happy to produce clumpable inputs all the time, with no underlying behavior (including noise).
I'm sure there are also tons of interesting models of human behavior that you can make out of some set of observations of human behavior via clustering. But just because you feed some data to an unsupervised learning algorithm and it discovers features doesn't imply that those features have any useful descriptive power to help us make sense of the natural world. THAT's anti-scientific thinking.
So if a causation turns out not to exist then finding natural patterns was an unreasonable starting point? What would have been a reasonable starting point in that circumstance?
So based on the study: http://advances.sciencemag.org/content/2/8/e1600451.full
and looking at the results in: http://d3a5ak6v9sb99l.cloudfront.net/content/advances/2/8/e1...
the summarization of "Undefined" as "Decides randomly" seems a little wrong- there's an obvious tendency in PD (Prisoner's Dilemma) for some results: http://d3a5ak6v9sb99l.cloudfront.net/content/advances/2/8/e1...
Also, I can't help but wonder whether there is bias effect, as humans seem to tend to use a small number of groups or factors for personalities. For example, in literature, J.K. Rowling's sorting hat chose between Gryffindor, Hufflepuff, Ravenclaw, and Slytherin, and Veronica Roth had citizens of Chicago choose between Abnegation, Amity, Candor, Dauntless, and Erudite. Even though there is a mix between races in humans, we tend to categorize- like in the U.S., typically you must choose from Caucausian, Asian-american, African-american, Pacific-islander, or Hispanic, even though color and genetic makeup vary. Then MBTI has four factors (E/I, S/N, T/F, J/P). Friedman and Rosenman came up with A and B types then later Denollet added the D type. We tend to categorize like this.
Also, they chose just four games for the study. That could have affected the outcome.
She tried so hard for ABCDE, but couldn't find the word "Benevolence" ?
It's sad that there is so much good social/psychological science out there, but people know so little about it and fall for this kind of ignorant reporting.
That model present five dimensions, to which the personality traits are mapped. It does not mean they are the basic dimensions of personality. Similarly I can map the world into 2 dimensions as in a photograph.
But I cannot say I've mapped the world to it's basic dimensions. I've lost quite a lot of data already - I'm missing 4 dimensions of data. The 3.rd coordinate, and 3 for momentum.
I actually think this study is pretty interesting. Full text here: http://advances.sciencemag.org/content/2/8/e1600451.full
I don't believe in their clustering, but then I rarely believe the output of any unsupervised learning methods.
Its an interesting approach, the real question is whether or not it captures incremental variance relative to an OCEAN (or the six factor one they like in medicine) baseline.
I agree it's interesting, but my original impression was also true: this isn't about discovering basic personality types. It's about correlating play across four 2 x 2 experimental games. This is a very, very narrow domain of social behaviour. Standard personality psychology, by contrast, aims to predict behaviour across a much wider range of situations. (I still think that is true, even if individual correlations are low in any one situation. But, if you know better and personality measures actually a crock of crap, feel free to correct me!)
Of course, it's possible that there really are four deep types of human personality, that you can capture them in these four games, and that they predict behaviour in lots and lots of domains. But I doubt it, and this article provides no evidence for it.
What's more interesting is the comparison between this unsupervised clustering method, and more theory-driven ways of categorising play in games, like inequality aversion models. I think there is an interesting race between psychologists, who pay greater attention to internal validity, and experimental economists who focus on theory consistency. I'd like to know which kind of models does better at predicting out of sample.
What is wrong with questioning existing knowledge, or proposing a completely different approach?
Who needs to know more than that we are a balance of sanguine, choleric, melancholic and phlegmatic? And that this balance is cause by precious bodily fluids, blood, the two biles and phlegm.
Bonus feature! You can fix any persons imbalances by simply letting the relevant fluids.
It sounds like it functions as another, parallel model to the one described in the article. Where's the harm?
After carrying out this kind of social experiment, the researchers developed a computer algorithm which set out to classify people according to their behavior. The computer algorith organized 90% of people into four groups:
It's not magic, innit. If you choose a bunch of features and run some classifier on them you'll get some grouping accoring to the distribution of values of those features in your sample. If you choose a different bunch of features, you'll get a different grouping. What did you learn? That your features have a distribution- well done.
Trying to clasisfy peoples' personalities is futile: what constitutes "behaviour" or "personality attributes" is very, very arbitrary and there's nothing to say it's not all in the eye of the beholder to begin with.
They are divergent :-). I found the characterization of "Envious" as people who always want to win a bit more charged than "Competitive" would have been.
> the latter of the four types, Envious, is the most common, with 30% compared to 20% for each of the other groups.
See also: Why are Adults so busy? https://news.ycombinator.com/item?id=12494999
I don't think this categorization is gonna get popular enough to hit the tabloids (you won't be filling up quizzes to see whether your significant other is 'Envious' or not), but it might be relevant to social science.
It could already be for making simulations using agent based models with "realistic" hypotheses by feeding in the strategies and the frequencies corresponding to these personalities into the model. Could be useful to social science research already.
Okay, it may turn out that this categorization is no better than Humourism, which seem to have suffered of the issues of poor variable operationalization as you described: ancient doctors frequently heaped together distinct phenomena ("hot" meant anything which was hot to the senses, whether it was a glowing hot iron, a flesh wound or a bowl of Jalapeños). But I don't think it's quite the case here: they're grouping users by their choices in a game. That's well defined enough, isn't it?
The names don't arouse any suspicion to me, either. I think it's just a case of researchers trying to "sexy up" their findings. It's something that might happen almost involuntarily: either because they started with a sexier model in their minds, or because in seeing their findings, it just came up to their minds and stuck there. I'm actually rather hard pressed for real examples right now, besides calling the transfer of a quantum state "teleportation" (because I just readn about that today), but it's something that happens quite often.
But that's probably just the INTJ in me talking.
1. The Envious (anything to get ahead of everyone)
2. The Optimists (everyone else is basically good, have faith)
3. The Pessimists (most others are evil, favor the lesser of evils)
4. Trusting Collaborators (will go along with anything to just be included)
5. Unclassifiable (no common archtype correlating decisions, strategic decision making demonstrates conflicting tendencies)They were not Gaussian distributions :)