[Book thoughts]: M. Gladwell’s Blink, Data modeling, and Corporate Structure

December 13, 2014 § 4 Comments

I got around to reading this book from my house a couple of weeks back as I have become increasingly intrigued by the art of decision making. This is an interesting discussion on the art of snap judgement, what the author calls ‘thin slicing’. There are several interesting and well written anecdotal evidence of instances where this technique is effective, misleading and not-applicable.

On the onset, I am not a fan of the Gladwell’s writing, in fact his entire approach of presenting a theory makes me uneasy and I will briefly explain why. Most of the socio-anthropological topics/theories he writes about are, in my opinion, to be thoroughly examined by scientific approaches, with rigor and thoroughness. The evidence must be collected without bias and examined skeptically. Any insights developed from observations must done so with caution.

Extraordinary claims require extraordinary evidence – Carl Sagan

Consider how Gladwell introduces John Gottman, a professor from University of  Washington working in relationship counselling and social sequence analysis, to build the case for the thin slicing theory.

John Gottman is a middle-aged man with owl-like eyes, silvery hair, and a trimmed beard. He is short and very charming, and when he talks about something that excites him his eyes light up even wider. During the Vietnam War, he was conscientious objector, and there is still something of a 60’s hippie about him, like the Mao cap he sometimes wears over his braided yarmulke.

I understand that the elaborate description of the person helps to perhaps build context, to prepare the reader for an outlandish claim that Gottman is about to make. To be fair, when reading a journal publication or before attending a technical talk, we have all scrolled down to the author’s photo and description or looked up the speaker’s homepage to get a sense of what to expect. But it belays a fear that what is about to come must not be taken too seriously either. Clearly, the author is carried away by the person’s appearance and credentials to sufficiently scrutinize what he is being presented with.

The book has, however, held my interest. This unformed style of writing has compelled me to make my own interpretations of the facts and theories presented to me. There is something Dr. House-esque about the situation, if I may say so. Thin slicing fits in with an important adage in the machine learning community, known as the Occam’s razor principle. The degraded version of the principle implies that simple works best. In this age of large data, and its unending sources of evidence to make decisions, this thumb-rule implies that plugging more information to your data model can infact be detrimental to its performance.

Imagine there are a large number of fuzzy input parameters that you could use to make a decision. In case of Gottman’s experiment, it is a video of a couple who are arguing on an issue and then go on to reconcile. If you were asked to predict if this couple would go on to have a lasting marriage, there are several things you could consider. For instance, you could endlessly profile each member, everything from who they are, what they do, likes, dislikes, attitude, personality, background, profession, interests. Then there are interactions, their behaviour towards one another, the words they used during confrontation and reconciliation. Perhaps you want to factor in what kind of day they had so far, even stress at work or health. To put it simply, your data model here can always seem incomplete, since to some extent at the least, there are always going to be factors that you are not privy to. Gottman concluded from his study however, that as little as four indicators of negative behaviour (criticism, contempt, defensiveness, and stonewalling) that can predict divorce with an accuracy of over 90%! Now, I for one can’t think of a fifth negative behaviour myself, so anyone with four of those going on in their lives seem pretty screwed.

This ofcourse is just one example and there are other interesting instances you could look at too. For instance, the car dealer who resisted pre-judging his customers as they entered his showroom. I often wonder about this contradiction of information, where less is more. It is something you see everywhere. Corporate hierarchy too in interesting in that sense, since the person who makes the biggest decisions, is the one with the least amount of ground-reality information! And to its credit, it works!

Bosses can be helpful!


Tagged: , , , , , , ,

§ 4 Responses to [Book thoughts]: M. Gladwell’s Blink, Data modeling, and Corporate Structure

  • Paridhi Jain says:

    Well, my limited knowledge can’t be used to comment either about the book or the corporate hierarchy but Occam’s razor principle. “Simple works best” or “less is more” models did not work for me so far. I always added information to reach closer to the best decision or the required judgement. Most people believe that more information to the (human) systems confuses them, and therefore, should be discarded. But, I think we fail to “fit” the “more” information in the existing systems and therefore, we fail.

    Edward de Bono in his book on “Lateral thinking” says that we often form patterns depending on the sequence of information we get. With new information, we need to restructure the patterns to give way to new patterns and make sense of the new information. I think we fail at rewiring the patterns. That is what confuses systems or humans. So, more is not bad, simple does not always work the best. Re-thinking is what is required.

    Liked by 1 person

  • samarth says:

    Very nicely put P! I agree with you, data analytics must always look for new and complimentary cues to make better decisions.


  • swarnabala says:

    it is very nice. ya……. a description of a person in depth gives the overall profile of the person and helps one to decide as to what to expect from the person. however, also remember many a time it so happens……………. it could be deceptive too………… “pappu lo kaalesaav” means you got it all wrong……………. that is also possible.. but from what i saw in my life………….. it is a very small percentage of instances in life…….. rare though………

    so is the reason, i guess, authors (at least the older ones) describe the entire road, the path, the sky and the earth around…………. before they take you into the village or a person they are going to introduce…… and then they again describe the person in-depth……….. “kallaki kattinatlugaa” . i am remembering what Chaganti Koteswara rao garu does….. he vividly describes the entry to exit in Tirumala tirupati………….. i think i also read the book………… if i see i will remember……………


  • Paridhi Jain says:

    Well, I was re-reading the “Occam Razor’s principle” and I missed an important part of it. “Simple is best” is decided for two competing theories which resulted in the same experimental results. It is like doing a feature importance analysis to know if we really needed to find all N features or k were sufficient to get the same results. Therefore, for me, the principle never applied, since with more data, I got better results.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

What’s this?

You are currently reading [Book thoughts]: M. Gladwell’s Blink, Data modeling, and Corporate Structure at In Theory....


%d bloggers like this: