cognitive biases 3 n.
Skip this Video
Loading SlideShow in 5 Seconds..
Cognitive Biases 3 PowerPoint Presentation
Download Presentation
Cognitive Biases 3

Cognitive Biases 3

178 Vues Download Presentation
Télécharger la présentation

Cognitive Biases 3

- - - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript

  1. Cognitive Biases 3

  2. Seeing what we expect to see

  3. Bias Our expectations often influence how we evaluate claims and evidence. We easily accept as true those things that we expect to be true, but are much more skeptical about things that are unexpected.

  4. Bias Bias can be a good thing. If someone tells you they saw a construction worker, it makes sense to believe them– construction workers are numerous, and we expect there to be numerous sightings of them. But if someone tells you they saw an extraterrestrial, things are different. You’ll be right to be skeptical: that is very unexpected.

  5. Bias Bias can also be a bad thing. If you’re biased against black people (“blacks tend toward criminal behavior,” you think), then you might be more likely to accept a negative statement about someone who is black, and more skeptical of believing positive things about them– even if they’re totally innocent, wonderful human beings.

  6. Confirmation Bias Last time we talked about confirmation bias: people are inclined to look for evidence that supports a hypothesis, and to ignore evidence that goes against it. This is important because the negative evidence (the evidence against a claim) is just as important as the positive evidence in evaluating the claim’s truth.

  7. Loftus & Palmer 1974 How things are described to us can affect how we see them. In one study, subjects were shown pictures of a car accident involving multiple cars. They were asked: “About how fast were the cars going when they (hit/smashed/collided/bumped/contacted ) each other?”

  8. Loftus & Palmer 1974

  9. Loftus & Palmer 1974 Additionally, the subjects were asked one week later whether they remembered seeing broken glass (from the cars) in the photographs. There was no glass, but subjects who had been asked “smash” or “collided” questions were more likely to remember some than subjects asked “contacted” or “hit” questions.

  10. Context Affects Expectation This study shows that context (how a picture is described to us) can affect how we see a thing (the picture itself), and what we remember about it.

  11. Studies have shown that people in many cultures have negative associations with the color black. They are biased against black-colored things.

  12. Frank & Gilovich 1988 One study asked professional referees (for American football) to watch a video clip of a play and decide whether the players deserved a penalty. In one version of the clip, the players wore white; in another, their uniforms were changed to black.

  13. Frank & Gilovich 1988 The referees were more likely to say that the players deserved a penalty if they were wearing black. Frank & Gilovich also found that teams with black uniforms actually did get penalized more often than teams with other colored uniforms!

  14. Disconfirmation Bias Our biases can lead us to accept evidence that agrees with our views and reject evidence against our views, even when the “for” and “against” evidence is of the same quality.

  15. Lord, Ross & Lepper 1979 One study looked at how people who were in favor of the death penalty evaluated arguments for and against it, and how people who were against the death penalty evaluated those same arguments. There were four types of arguments:

  16. Lord, Ross & Lepper 1979 AGAINST-SAME. A study that showed that murder rates in a state increased after that state instituted the death penalty. AGAINST-DIFF. A study that showed that murder rates were higher in states that had the death penalty than in states that didn’t.

  17. Lord, Ross & Lepper 1979 FOR-SAME. A study that showed that murder rates in a state decreased after that state instituted the death penalty. FOR-DIFF. A study that showed that murder rates were lower in states that had the death penalty than in states that didn’t.

  18. Lord, Ross & Lepper 1979 All the subjects got one study AGAINST the death penalty and one study FOR it. If the study they got AGAINST it was the SAME condition, then they got FOR-DIFF; if the study they got AGAINST it was DIFF, then they got FOR-SAME.

  19. Lord, Ross & Lepper 1979 People who liked the death penalty and received AGAINST-SAME and FOR-DIFF argued that SAME studies were bad, and DIFF ones were good. They liked the study that supported them. If they got AGAINST-DIFF and FOR-SAME, they argued the opposite: that DIFF studies were bad, and that SAME studies were good.

  20. Lord, Ross & Lepper 1979 The same was true for people who opposed the death penalty: they liked SAME studies when they got AGAINST-SAME, but not when they got FOR-SAME; they liked DIFF studies when they got AGAINST-DIFF, but not when they got FOR-DIFF. Everyone liked the studies that agreed with them!

  21. Lord, Ross & Lepper 1979 What’s interesting is that the arguments given by the subjects about why DIFF studies are bad (or why SAME studies are bad) were good arguments. No one was arguing in bad faith. But their biases made them see the flaws in studies that disagreed with them, and made them ignore the flaws in the studies that agreed with them.

  22. Disconfirmation Bias This is sometimes called disconfirmation bias. It is the tendency to subject evidence against your views to a greater degree of scrutiny than evidence in favor of your views. It is a double-standard for evidence evaluation.

  23. Gamblers A similar study was seen in a study of gamblers (in particular, ones who bet on sports matches). Why do gamblers keep gambling when they lose money so often? Don’t they ever learn from their mistakes? Are they just forgetting about all those times they lost?

  24. Gamblers The study showed that gamblers didn’t forget their losses– they were more likely to remember them. What they did was explain away the times they lost. Losses were counted as accidents, bad luck, “near wins”, etc. They weren’t counted as bad gambles or mistakes.

  25. Gamblers Wins however were never attributed to accident or luck. They were the result of skill. So wins and losses were treated differently. A win was taken as evidence that the gambler was good at choosing bets. But losses were not taken as evidence that the gambler was bad at choosing bets, just that he was “unlucky”.

  26. Evidential Double-Standards Disconfirmation bias and evidential double-standards can play a big (negative) role in political disputes. As a recent example, Harvard Law Professor Alan Dershowitz has criticized Brooklyn College in New York for hosting an event with speakers critical of Israel, but no opposing voices.

  27. Double-Standard That might be a good reason to object to an event, that it only presents one side. But Dershowitz is applying a double-standard. Dershowitz himself has spoken several times at Brooklyn College, and never once had anyone else speaking against his views at those times, and never requested anyone to!

  28. Multiple Endpoints One way that our expectations can influence our beliefs is called “the problem of multiple endpoints.”

  29. Multiple Endpoints For example, suppose someone claims that spending time on Facebook is bad for your social life. Well what does that mean? There are lots of effects of Facebook that might be counted as bad for your social life (spending less time face-to-face with friends) and lots of ways that might be counted as good (talking more often with a larger range of friends).

  30. Multiple Endpoints Here, if I’m biased because I dislike Facebook, I can interpret “bad for your social life” in a way that Facebook is bad; if I like Facebook, I can interpret it in a way that is good. There are multiple different events (“endpoints”) that I can focus on when evaluating the claim.

  31. Multiple Endpoints People often say to parents, “your child looks so much like you!” even if (unknown to them) the child is adopted. There are lots of ways that any two people can look alike (hair, eyes, mouth, nose…). Most people will be similar in some ways. If you expect that two people will look similar (you think they’re biologically related), you will focus on those “endpoints”.

  32. John Edward So-called “psychics” often use multiple endpoints to their advantage to convince people they have supernatural powers.

  33. Lots of People with Relations In this clip, Edward asks if there’s a son, younger son, grandson, or younger brother who “passed.” Surely in a large audience some person is related to some one other male who died at some point in the past. It could be a long time ago, or recent. It could be a son, brother, grandson…

  34. Vague Guess Sometimes Wrong Edward guesses that this person died in “an event.” Can you think of someone who died young whose death wouldn’t be an “event.” It can’t be natural! He insists that something is related to June. And statistically it’s a good guess: something usually happens every month. But here he’s wrong.

  35. Last-Ditch Effort Finally, not looking too good, Edward quickly guesses two very likely things: that someone in her family had “brain cancer or something brain-related.” How many old people have not had something brain related? Second, he guesses her dog died. Lots of Americans have dogs, many have many dogs, and dogs don’t live all that long.

  36. Main Point The main point is this: very strange claims like “dead people can speak to John Edward and pass on messages to him for the living” can seem to be supported by the evidence. But this is only because very vague claims that can be made true by multiple endpoints are almost always true. They don’t support anything.

  37. Two-Sided Events Some events are “one-sided” and some events are “two-sided”. A “two-sided” event is an event that can turn out one of two ways. For example, if I bet on a football match, I can either win or I can lose. Winning is one “side” of the event and losing is the other “side”.

  38. One-Sided Events “One-sided” events are different. They can either (i) turn out in exactly one way or (ii) not exist.

  39. One-Sided Events For example, suppose I believe that the phone always rings when I’m in the shower. If I’m in the shower and the phone rings, that seems like an event– something happened: I was in the shower, and the phone rang.

  40. One-Sided Events For example, suppose I believe that the phone always rings when I’m in the shower. But if I’m in the shower and the phone doesn’t ring, it seems like nothing happened. I don’t say “look what happened when I was in the shower– the phone didn’t ring!”

  41. Two-Sided Events & Memory For two-sided events, we notice both outcomes equally. So when evaluating claims about them, like “I always win when I bet on a match,” we will be confronted with the negative evidence (all those times I bet on a match and lost).

  42. For one-sided events, we only notice when they happen, not when they don’t happen. So it may seem to me that my phone rings frequently when I’m in the shower, because I never notice the many, many times it does not ring when I’m in the shower.

  43. Confirmations and Non-confirmations Sometimes events are “one-sided” because we are never inclined to notice the other side. Only when the event confirms a certain belief do we notice its relevance to the belief; if it doesn’t confirm the belief, then we’re uninclined to notice its relevance.

  44. Confirmations and Non-confirmations Suppose a fortune teller predicts that I will have twins. If many years later, I have twins, I may remember the prediction. I have evidence that fortune tellers are accurate. If instead I have only one child, I’m less likely to remember the prophecy. One child doesn’t remind me of a prophecy about two children.

  45. Open-Ended Claims Sometimes events are “one-sided” because the claims they are relevant to are open-ended. For example, suppose I predict that you will someday break a bone. If you break a bone tomorrow, next week, next month, next year, 30 years from now, etc. then the relevant event happened. I was right.

  46. But if you don’t break a bone tomorrow, nothing happened. If you don’t break it next week, nothing happened. If you don’t break it next month, nothing happened. None of this shows that I’m wrong. Only if you died without breaking a bone would I be wrong. And you wouldn’t be around to notice that.