Ekman's Telling Lies

[Non-fiction/Psychology and emotion] [***]

I read Paul Ekman's Telling Lies hoping to be exposed to more Ekman's groundbreaking research on human expression of emotion and how that emotion gets expressed and concealed in what people say. The book covers this material well, but unfortunately it has many other deficiencies that detract from what it does well. In particular, Ekman dabbles in philosophy, politics, and ethics unnecessarily, and without any of the precision of thought that I would expect from a book on one of those subjects, and throughout Telling Lies makes arguments that display a basic lack of understanding in relevant statistics.

Though Telling Lies is a chapter book that was clearly intended to be read (and probably should be read once by anyone who wants to learn from it), the book makes for an excellent reference and probably serves in this capacity better than it serves as an informative read. The appendices at the end of the book provide a clear summary of the books most useful information, and a 15 page index makes navigating the 300 pages of the books primary content relatively easy. (Approximately 80 pages of the 120 pages of bonus content added to this edition are not indexed, but they don't contain as much useful information that would be worth referencing again later as the main part of the work that is thoroughly indexed.) For anyone to whom it's worth reading, it's almost certainly worth keeping on hand to periodically consult as well.

The book covers how people's emotions leak into their body language and facial expressions even when they are trying to prevent such leakages, and also how the level of mental agreement the speaker has with what she is saying impacts her use of gestures and other cues. It also covers the general practice of lie detecting and discusses how the average person, and even the average police officer and psychologist (who are in the business of detecting lies) do no better than chance, on average, at detecting whether or not a speaker is lying. Polygraphs do slightly better, and some people who by nature or training have learned to focus on the right cues do much better still. Interestingly, while most people consider themselves to be excellent lie detectors (and thus unlikely to consider others to be significantly better), people who interact with each other regularly naturally notice who among them are actually adept at catching lies, and when police officers are interviewed to determine who the best interrogators in their station, they successfully identify the few members of their force that do significantly outperform chance at detecting lies.

If there was ever a book that should have used Bayesian statistics significantly, Telling Lies was it. Practically every topic that Ekman covers involves Ekman struggling to convey in words intuitions he has learned from a lifetime of studying lie detecting that would be perfectly captured by statistics, if Ekman were more familiar with that formalism. I would love to see this book rewritten from a Bayesian perspective with a few other minor problems corrected. Its content could make for a fantastic book, but unfortunately Ekman struggles to present it well.

If you are familiar with Bayesian statistics and used to the idea of updating the probability you assign to each of your possible theories in accordance with how strongly each theory predicts each new observation, then you can safely read this book as long as you continuously make the appropriate mental substitutions for whatever it is that Ekman says about proof and evidence in any given place. All of his many statements of the form "if you see X, it is evidence in favor of B, but if you don't see X it is not evidence against B" are simply inaccurate from the point of view of statistics, and should be replaced with the appropriate Bayesian inferences. Interestingly, Ekman is aware of the inadequacy of his view of evidence, on some level or another. In the process of explaining his views on polygraph testing, he finds himself needing a more nuanced view of evidence than he uses throughout the rest of this book, and does eventually put forth his argument using accurate statistics. Readers unfamiliar with the basics of Bayesian statistics (and unenthusiastic about the prospect of learning) should start with chapter seven, and then return to the rest of the beginning bearing in mind that his illustrations and explanation about how the base rate of lying in the population affect the percentage of false positives and false negatives in the polygraphs results applies equally to everything he says about evidence because it is a general feature of statistics, not a particular feature of polygraph tests.

One more thing that I think deserves mention about this book is that Ekman draws many of his examples from fiction and most of his "real life" examples are drawn from historical events which he himself never witnessed. At first, I considered this a weakness of the book, but upon further reflection I think that it is simply a decision of presentation, and moreover, one that works well. Though he uses fictional evidence to support and illustrate his points, Ekman isn't generalizing from fictional evidence. He formed his beliefs through decades of research, most of it high quality research with strong results. Since he legitimately derived his opinions by studying actual human behavior, I don't see anything wrong with him using fictional evidence in presenting his beliefs and findings to a wider audience. Rigorous presentation of data *might* be more persuasive to the few people who have conditioned themselves to respond to it appropriately, but the average reader is more likely to be convinced by a story, even a fictitious one (especially since the average reader is unlikely to even read a book that relies on something more rigorous).

On the whole, I'd recommend Telling Lies, at least to some people some of the time. It's not perfect, but the positives outweigh the minuses for anyone who approaches it intelligently.