Post by Ray Poynter, 22 October 2019
Abstract: Ray Poynter reviews Malcolm Gladwell’s new book Talking to Strangers. Ray’s recommendation is to read the book (or listen to it on Audible), but not to accept everything you hear as science or fact.
I’ve just finished listening to the Audible version of Malcolm Gladwell’s new book Talking to Strangers. I’d love to hear what other people think of the book?
I like what Gladwell and his publisher have done with the Audible format. Rather than read the written text it’s full of actual news recordings, recordings of police interviews, and of actors playing the parts when dealing with things like court records (if a recording of the actual event is not available). This uses the audio format is far better than just reading the text. It offers something the text does not offer.
My top takes from the book are:
- Gladwell’s science is often a bit dodgy (for example, the Tipping Point oversimplifies and it misses issues such as homophily), so I will cross-reference the key points before I fully take them on board, rather than simply assuming Gladwell is right.
- The key message in the book is that in many situations we are bad are reading other people. This is not a surprise, but an important reminder for market researchers.
- One of the key reasons that fraudsters, imposters, spies etc often do so well is that human’s are pre-wired to believe people, it is the default setting. When fraudsters are caught, Gladwell says, it is often because they have run into somebody who does not have the normal wiring. However, Gladwell makes the point that human society benefits massively from being biased towards belief, it makes the world a better place, 99% of the time – life would be quite grim if disbelieving was the default.
- One interesting example of misreading people is when Gladwell looks at the performance of judges compared with an algorithm. The algorithm only had things like demographics, previous record, and the crime details, but it was a better predictor of risk than judges who had that data AND got the chance to meet people and ask them questions. It is suggested that judges made WORSE decisions because they meet people and the judges think they can read people.
- Specialists at interpreting body language and faces (e.g. trained investigators and interrogators) do better than the general public when the messages are not mismatched, but they do worse than the general public when there is a mismatch. A mismatch is, for example, when an innocent person who is nervous is judged to be guilty of something. Another mismatch is when a guilty person who is confident – Gladwell cites the case of Bernie Madoff, where the data made it clear that this was a scam, but he convinced almost everybody he met that it was OK, for years.
- The book suggests that many terrible recent mistakes were more understandable than they at first appeared, e.g. the arrests Amanda Knox and Sandra Bland. In both of these cases, there was a mismatch between the observed behaviour and how the police interpreted it.
- Another theme that runs through the book is how often a finding from one situation does not translate to other contexts. Gladwell spends quite a bit of time looking at policing initiatives that worked in their first trial, in a specific location, with specific rules, but which were disastrous in other locations and contexts.
I definitely recommend this book, it is well crafted and the Audible version is very easy to listen to (except for some of the detailed bits about crimes and torture). The book does not provide any answers (IMHO), but it provides lots of great questions.
Two interesting questions for market researchers are:
- In qualitative research, most practitioners prefer face-to-face techniques (as opposed to, say, online chat) because they can interpret the body language and tone of the participants. However, Gladwell’s case is that most experts in interpreting people are actually much worse at it than they think. Perhaps we should conduct more research-on-research to find out which produces the ‘best’ results, not which appears best to the people involved.
- In quantitative research, there is a growing discussion about the role of AI in interpreting data. Researchers tend to be a bit smug that humans are better at understanding the why and finding the story (compared to machines), but is that necessarily so?
This assumption about the power of humans reminds me of a great webinar presentation that Janneke van den Bent from SKIM did for #NewMR last September. Janneke reported on a project where SKIM used two teams, one pure human, one AI assisted. One of the key findings was that the end client preferred the AI+Human report to the Human only report, it was clearer and easier to digest. Perhaps this is a market research-related example of what Gladwell is talking about, our tendency to over (and incorrectly) interpret. To watch a recording of Janneke’s presentation, click here.