Duped: Book Review
Duped: Truth-Default Theory and the Social Science of Lying and Deception (2020), Tim Levine. Gladwell's Talking to Strangers used truth-default theory and noted on Twitter when this book came out. Lots of useful content, although a somewhat annoying read. Levine focuses a lot on other deception theories he believes are wrong and other issues on the general field. I'm interested in his research and perspective, which more or less starts with Chapter 6. He points out how rampant "fake news" and related issues have cropped up especially since 2016 and the election of Trump, pointing out that people can be gullible. People generally have a mind-set he calls "truth-default," uncritically accepting messages as honest. He notes that blurting out incorrect statements viewed as authenticity and confident interpreted as honest. But speaking like a politician with carefully chosen words is not to be trusted (ergo, Clinton). The problem is that appearing honest and being honest are unrelated. Truth default theory (TDT) means communication uncritically accepted as true: "an adaptive product of human evolution that enables efficient communication and social coordination" (p. xi). He also notes the existence of a few prolific liars. Unlike other deception theories, TDT involves listening to what is said, not how it is said; using evidence and skilled questioning produces better outcomes than observing deception cues like gaze aversion, facial expressions and so on.
Part I is on deception in the social sciences, which Levine mainly critiques and dismisses (skipping to Part II not a bad idea). "The best lies are never detected" (p. 3); plus liars are quite different fro each other. Behaviors noted to give away liars misclassify honest people. Lots of clues come with hindsight, rather than on the spot. Examples used in experiments are cherry-picked with hindsight information. Important questions are: what people look for determining if someone is lying; what are distinguish truthful communications; how accurate are people distinguishing truth from lies; what makes people more accurate. Doubt comes from people being nervous, hesitant, uncertain or lack confidence. Honesty associated with friendly composed, confident, and engaged--these are not related to honesty. Experiments show about 54% accuracy, just better than chance. Levine noted that experiments often use deception (Milgram authority tests) and confederates, but subjects almost never detect this deception. Chapter 2 is on cues, what people look for to determine is someone is lying, with failing to maintain eye contact (gaze aversion) number 1 but over 100 cues investigated (jut have little validity). [List on p. 25.] Chapter 3: Deception Detection Accuracy. The first deception detection experiment was in 1941 (Fay and Middleton), with correct responses of 55.6% by "judges." Multiple deception theories sprung up identifying different deception cues, suggesting factors like longer duration interviews,sender motivation, training to be better lie detectors (54% to 58%), variance in senders more than judges. Chapter 4: Rivals. Six prior theories: Ekman's leakage theory, Ekman's updated, four-factor theory, Bella DePaulo's self-presentation perspective, interpersonal deception theory (IDT), and Vrij's cognitive load approach. Factors include micro facial expressions, fidgeting and other body movement, other nonverbal factors. Four factor: felt emotions, arousal, cognitive effort, and attempted behavio ral control. Mediators come in the middle. IDT says deception commonplace. Vrij focus on legal and criminal cases. Importance of interviewers, prompting cues (e.g., narrative in reverse chronology, unanticipated questions. Chapter 5: Critiquing the Rivals. Levine talks about good theory, which sounds good but the social sciences do not seem to do that well in designing theories that have the stated criteria (coherent, logically consistent, efficient, valid knowledge, prediction and explanation). At the top here are cue theories, which overall are not not very predictive (although cue effects exist at the individual study level, just not generalizable).
Part II. Truth Default Theory. TDT says we assume people are honest, which works most of the time (efficient communications), but makes us vulnerable to occasional deceit. Deception: intentional purposeful misleading statement. A lie is a subtype involving outright falsehood. Truth-lie base-rate: proportion of messages that are honest. Demeanor: intercorrelated behaviors or cues on how people present themselves. Honest demeanor: seen as honest. Suspicion: suspended judgment and uncertainty about honesty. Transparency: extent to which the communication is honest or deceptive. Coherence: logical consistency of content. TDT as a modular theory (various models and hypotheses within it). A few prolific liars ("outliars"). Deception motives: people have a reason to lie. Veracity effect: the honesty of communication predicts whether the message will be judged correctly; e.g., honest messages produce higher accuracy than lies. Park-Levine Probability Model: honest messages should result in higher accuracy, because of base-rates and veracity effect. Sender honest demeanor: how honest a person seems depends on 11 factors; honest demeanor has little to do with actual honesty. Few lies are detected in real time. Diagnostic utility involves using useful information; correspondence relates what is said to known facts and evidence (fact checking. Questions effects: asking the right questions to get diagnostically useful information. Expertise in deception detection highly context dependent and involves knowing how to prompt useful information. Deception is typically tactical; people are honest unless truth threatens some desired goal. Truth default requires a trigger event to abandon it: projected motive, behavioral displays, lack of coherence or correspondence, and information from a third party. Deceptions may be detected well after the fact. Demeanor-based detection only slightly better than chance. Accurate detection best through confession or external evidence.
Chapter 7: Defining Deception. Start with bald faced lies (BFLs) and bald-faced truths (BFTs) (Steve McCornack); deception as an intentional act to mislead others; a lie as sub-type as outright falsehood, known to be false. Honest communication lacks deception purpose, intent or awareness (not necessarily fully true or involve full disclosure). Sarcasm is not deception. Deception can be true, the key point is conscious intent to mislead. Levine: "deception involves knowingly or intentionally misleading another person" (p. 105). Note: Levine views strategic omissions or ambiguity is deception. Delusional messages are not deceptive. "Useful distinctions between conscious intentional deceptions, deceptive attempts, messages perceived as deceptive, and messages that are functionally deceptive" (p. 115). Chapter 8: Information Manipulation. Paul Grice, philosophy of language, logic versus everyday communication. Cooperation principle: quantity (make communication informative), quality (make it true), make it relevant and on topic, and be clear: statements that are informative, accurate, relevant and unambiguous. "Most deceptive messages involved some combination of omitted information (quantity violation), false information (quality violation), a lack of clarity functioning to obscure information (manner violation), and/or irrelevant information, providing a diversion of attention (relevance violation). Finally, existing taxonomies of deception types (e.g., omissions, half-truths, exaggerations, etc.) failed to capture the variations in deception messages" (p.120). "We get fooled because the violations are covert" (p. 120). Note importance of degrees of deceptiveness. [This seems especially relevant in political talk.] Message production: "People opportunistically start down the easiest path that seems to work, adjusting course as they go" (p. 127). "Deceptive intent is the result of a contextual problem-solving process where deception is deemed the most efficient solution" (p. 129). The most frequent approach to deception is omission. "When do people lie? When they judge the truthful information as being so problematic that it cannot be disclosed" (p. 132); basically people have things to hide.
Chapter 9: Prevalence. TDT is based truth-lie base rate; truth-default is adaptive. Estimating prevalence uses surveys, diary method, and experiments. Deception motives model based on the idea that people lie for a reason and deception is not random (and varies across people and situations); also, people deceive when truth is problematic. Deception is expected to be infrequent. Diary method suggests people lie one to two times a day (average 1.65), but skewed; "white lies" more common. Chapter 10: Deception Motives. Lying and deception may be predictable, because people lie for a reason usually for some other reason. Deception is reserved to situations where honesty would be problematic. Ethicist Sissela Bok holds that lying requires justification while honesty does not; a practical rather than a moral choice. Steve McCornack views deception as problem solving, where honesty interferes with some desired goal. Five TDT claims: 1. people lie for a reason (purposeful, not random); 2. deception is tactical, not ultimate goal but a way to achieve the goal; 3. motives for truthful and deceptive communication are the same (e.g., to avoid punishment--honest people and liars say the same sorts of things; be polite); 4. when truth is consistent with goals, a person will almost always be honest [problem of pathological liars]; 5. deception becomes problematic when truth makes honest communication difficult or inefficient. Cheating experiment: cheaters are more prone to lie. List of deception motives: save face; manage relationships' to exploit; avoid tension or conflict; and control situations. [Longer lists starting on p. 165.] Worst lies in Pakistan, 17% malicious, 43% involved fraud. Questions that make innocent people seem guilty and visa versa. Don't overlook the obvious lie: deception motives determine deception communication.
Chapter 11: Truth-Bias and Truth-Default. Most people are truth-biased most of the time. TDT follows truth-bias, as a cognitive default state. They are adaptive and can be switched off. Prior to TDT, truth-bias was viewed as an error reflecting flawed judgment. Levine states that both are functional , adaptive and improve accuracy. TDT as consistent with Kahneman-Tversky cognitive heuristics ("relatively mindless decision rules that people use to make intuitive judgments under conditions of uncertainty ... can lead to biased, less-than-optimal choices" p. 177); or they lead to okay decisions where they are used. Truth bias: "believe that another person's communication is honest, independent of its actual honesty" (p. 177). In experiments, moderate suspicion improved accuracy. Truth-bias is higher in face-to-face communication; importance of demeanor. A trigger catches our attention and can lead to suspicion; possibility of suspended belief/uncertainty or passive truth default. Absent prompting, truth-default is pervasive. Prompting makes a difference; e.g., trigger events, behavioral displays, lack of coherence. Assess honesty: context and motive, sender demeanor, information from third parties, coherence, and correspondence information (p. 194); plus message plausibility.
Chapter 12: The Veracity Effect and Truth-Lie Base Rates. Truth-bias would impair accuracy. As suspicion increased, truth-bias would decline and accuracy improve. Moderate suspicion would lead to higher accuracy than low or high suspicion. Suspicion reduced but did not eliminate truth-bias. Veracity effect: honest messages produce higher accuracy, resulting from truth-bias. Veracity effect is biggest when suspicion is low; the opposite for lies (lie accuracy). Veracity effect: truth accuracy minus lie accuracy. Veracity effect weaken with experts, when suspicious and senders have less honest demeanor. Truth-lie base rate: ratio of truths to lies in a lie detection task. People are insensitive to base-rates in the absence of a trigger. They become sensitive to base-rates when triggers are present and informative about actual base-rates. Park-Levine probability model: honest messages have higher accuracy (veracity effect). Chapter 13: Explaining Slightly-Better-Than-Chance Accuracy. Why this lousy result. People look for the wrong cues. Accuracy is better than chance because judges get transparent senders right; not much better because most senders aren't transparent. Truth training (note: placebo effect) reduces truth-bias (makes judges more cynical) but does not affect truth-lie discrimination. Senders vary in transparency and demeanor. Judges always accurate with sender transparency (e.g., transparent liars are bad liars; people without transparency are inscrutable; judges get wrong senders with negative transparency). Judges follow demeanor (tendency for senders to be believed independent of actual honesty). In the veracity effect, truth-bias makes judges right about truths but wrong about the lies. Four quadrants of demeanor: appear honest or dishonest, actually be honest or dishonest--[specific characteristics listed on p. 248.] Two quadrants are matched, two are mismatched (honesty and demeanor mismatched--sincere appearing liars, suspicious honest senders). Judges generally correct about matched senders and wrong about mismatched senders. Experts (government agents) were right about all matched senders, but only 20% on mismatched senders. This leads to believability quotient (demeanor index), honest demeanor=1. BQ scores predict judges honesty ratings, not actual honesty. High BQ ratings useful for politicians, attorneys, salespeople.
Chapter 14: Improving Accuracy. Something that contradicts prior knowledge; real-time judgments not the likely way to detect lies--most detected after the fact, either comparing what was said to external evidence or the liar later told the truth. Test: 15% of lies detected immediately, 80% later based on external evidence, 30% of liars confessed. Correspondence: extent to which what is said matches (corresponds) what known facts and evidence (if not, they are objectively false). Coherence: logical consistency. Fact checking is best bet for discovering truth. A false statement is not necessarily deceptive; it could be an honest mistake or person delusional. Messenger must know message is false. Context (background) information useful, improving judges accuracy. Diagnostic questioning, prompting and using useful information. Probing effect impacts believability: open-ended questions; "is there anything you haven't told me?" "Why should I believe you?" Last question can throw non-cheaters off, but cheaters may anticipate it. Expertise: (strategic questions) knowing how to question honest people seem honest and liars are revealed. Experts are good listeners, circling back and paying attention to logical consistency. Chapter 15: The TDT Perspective, people as passive believers. Solution to deception is deterrence, based on religion, cultures, and legal-political-regulatory structures, usually after-the-fact. Deception is not random, people do it for a reason. Understanding incentive structures and people's character is important (more than person's demeanor). Improved lie detection comes from attention to communication content (e.g., verifiable details), not cues. People usually lie on as as-needed basis.