Unformatted text preview: Eyewitness Identification Sources of systematic error (bias) Overconfidence Background beliefs about likely perpetrators Poor observing conditions (amplifies effects of overconfidence) Story reconstruction Lineups (number of times you see a face, etc.) 1 Base Rates and Error Rates Base rate information Was the cab yellow or white? What is the relative frequency of white cabs, or yellow cabs, in the population? The eyewitness identification has an error rate associated with it.
2 Overconfident Identification There are 2 outcomes for identification a. Identify the wrong person b. Fail to identify the right one Overconfidence leads to (a). Overconfidence in highly significant reasoning problems can lead to fatally inaccurate judgments for example, that guilt has been proven beyond a reasonable doubt. There have been 110 people who were convicted of murder who have been found to be not guilty on the basis of DNA evidence. Most of these errors have been found to be the result of mistaken eyewitness testimony and the overconfidence of eyewitnesses and juries in the reliability of such testimony (Wells, Olson, and Charman, 2002).
3 Some Background
For the defender of commonsense looking for a process that counterbalances the effects of the biases, you won't find it in memory. Memory is not like a videotape Recall is not like the act of playing the tape back. Remembering is reconstructive (Loftus). We fill in missing detail. (There is a common example used to establish this point. Think of an event in your past that you found pleasurable. Now, did you see yourself in the setting? If you did (and most people do), you must have reconstructed the event so that you were insinuated in the scene, because in the original event you certainly didn't observe yourself (Myers 1990).
4 Experiments on Eyewitness Reliability But our understanding of an event is affected by how it is cast, and the patterns of connection is suggested by the cast. In one famous experiment, Loftus and Palmer (1974) asked 45 students to view seven different film clips of a car accident. They were divided into 5 groups. One group was asked "About how fast were the cars going when they contacted each other?" For each of the other groups, the word "contacted" was replaced respectively with "hit", "bumped", "collided", or "smashed". 5 The group that received the word "smashed" assigned a speed 9 miles per hour faster than that receiving the word "contacted", leading Loftus and Palmer to conclude that even small differences in the specific form of the question can guide the reconstruction of the event. In a second experiment, Loftus and Palmer found that the question "About how fast were the cars going when they smashed into each other?" yielded the highest proportion of answers that the accident included broken glass, even though the film clips involved none.
6 Results Where do you set your criterion? How certain do I have to be before I say "He did it"? How sensitive am I to the most diagnostic cues? If insensitive, then you are fishing with a large mesh you won't catch many but the ones you will catch are big If sensitive, you are fishing with a small mesh you catch many, including ones you weren't supposed to or that you didn't want to You could ask these 2 questions as a matter of policy That's what parole boards do Gary Wells is the person who is most know to do psychological research in line ups 7 "Simultaneous" lineups At present, eyewitnesses browse through photographs of suspects, comparing, contrasting and restudying them at will. Sources of error: seeing different faces different numbers of time, with no control on order. This produces familiarity effects. But familiarity is not reliably related to accuracy. You can study them without anyone interfering Familiarity effects are known as fluency effects Produce good affect and that means that there is positive affect with fluency The process tends to be unreliable 8 The right sized mesh Sequential viewing of photographs has been shown to cut down on the number of false identifications by eyewitnesses without reducing the number of correct ones . You can test this in cases where you know who the actual perpetrator is. Sequential viewing there is a specific order that you have to go through and if you stop for any reason you have to go through the rest If you look at one face you HAVE to look at them all If you can increase hits and decrease false alarms that is the ideal situation The next best thing is to reduce the number of false alarms without increasing hits You want a high hit rate and really low false alarm rate You don't false alarms high without being able to do anything to the hits You want a hit rate as high as possible but not if that requires high false alarms 9 Sequential lineups The difference between the old and new systems is subtle but highly significant, according to researchers who have studied the psychology of witness identification Under the new system, victims and other eyewitnesses would be shown pictures one after the other. They would not be allowed to browse. If they wanted a second look, they would have to view all the photos a second time, in a new sequence. Also, the pictures would usually be shown by a person who would not know who the real suspect was. People are very driven by startup costs 10 Outside strategies for debiasing
The sequential lineup is a good example of controlling the mind by controlling informational input, an outside strategy. An inside strategy is a debiasing strategy that uses little mental exercises Outside strategy control the bias like putting cones up They are changing the conditions that are prone to mistakes Doesn't ask for a mental exercise from the decisionmaker Most policy solutions are outside strategy
11 Policy and implementation issues The relation between research and policy Concerns expressed in the Wells article What does the research show? Presumably you reach a ceiling because certain kinds of limits are reached Wells was sure that this would be highly opposed by police departments You can take ordinary citizens and improve their judgement Hard to predict What will the public, and policymakers, accept? 12 Predictive accuracy and fairness Monahan article Looks at the predictive accuracy and fairness You can make it more accurate by adding things to it that the human will not be able to predict What happens when greater accuracy can only be achieved by using demographic categories over which the individual has no control: gender, race, etc. 13 Further Readings Elizabeth Loftus Eyewitness for the defense Anything by Gary Wells. For examples of your own tendencies of eyewitness misidentification, see his website: http://www.psychology.iastate.edu/faculty/gwell s/homepage.htm 14 Common Hard Choices If we didn't sacrifice accuracy for efficiency then why aren't there more protected values that we hold on to "I would never sacrifice accuracy for efficiency" Triage and Meatball Surgery If so, then why are there so few "protected values" (e.g., education, medical care, road work, water services, etc.)? Wait lists (refer to psychological services that you are put on a wait list for), high caseloads, and short time limits for indigent psychological services, legal services, etc. When you are dealing with limited resources 15 Risk and Race Two goals of good reasoning strategies: Accuracy and Efficiency Accuracy = closest to the truth Efficiency = the least costly given the goal A strategy that gets you the closest to the truth Promotes beliefs that are closest to the truth You may need to trade them off at a certain point
16 Barely satisfies the minimum requirement The cold, hard truths So, we DO sacrifice accuracy and other epistemic virtues (virtues having to do with knowledge) for low cost or high volume. Can we do it in the case of justice? The answer to this question depends on the goals of a system of justice. We need to process so many people that it would be useful to have an efficient system EX: RETRIVITIVISM =the reason you should be punished is because you did something wrong and not because you will do it again 17 What is the law supposed to do? We will make such sacrifices, depending on what we expect the law to do. Punish Criminals to give them just deserts? Punish Criminals to reform them? Protect society? 18 Protecting Society If protecting society were the only goal, or even the main one, then warehousing violent criminals would be fine with people. This might involve fluffy pillows, lots of TV, and plenty of recreation even if no freedom to leave. But such comfort, no matter how sequestered, is not ok with people. Ask them. But notice how much weight to put on the reaction to those people That kind of comfort in a warehouse setting is not ok TWO QUESTIONS Why does it bother people that warehouse criminals are comfortable? 19 Civil/Criminal distinction Punishing criminals vs. punishing psychiatric patients Psychiatric patients don't have the brains to intentionally commit a crime We don't want to punish psychiatric patients, just warehouse them You only want to punish the former class The interest of criminal retribution is often different from criminal civil protections People can be committed criminally or civilly We don't want to do the latter. Why? Criminal retribution vs. civil protection
20 Direction of Effects Just Deserts Backward looking blameworthy for past events What are just deserts? It is a retributivist approach. Crime control Futurelooking It focuses on reform Youre in effect getting retribution Warehousing You want to prevent stuff from happening in the future You are trying to protect society and the best way to do that is to remove criminals
21 Demographic categories of an SPR
Would the person commit criminal acts of violence that would constitute a continuing threat to society Race Gender Ethnicity the people who don't recidivate are those that this will happen to later in life 22 When can we use valid scientific evidence of risk factors? In criminal law, we should restrict use of violence risk factors to those that indicate the extent or seriousness of past criminal conduct. In civil commitment (for mental patients for example) we should allow assessments of risk of future violence, independent of past criminal conduct. Blameworthiness should play no part.
23 Any factors except.... Due to equal protection considerations, any factors that accurately forecast violence should be used except race and ethnicity. 24 Result from Will There is a principle at play here: Factors determining sentencing, etc., must be related to the what the agent did as an act of the will, rather than in virtue of unchosen membership in a demographic category. 25 A Fixed Past? Adults abused as children are far more likely to commit violent crimes than others. Can you use this as a risk factor? Once again, their abuse has little to do with their past CHOICES. 26 Why do we find it so hard to empathize with our future selves, or with present others. Psychic Numbing used to explain why people discount the way that they do
ANSWER: psychological distance Psychological Remoteness Geographical and Temporal Discounting Binding through Policy Avoiding Donor Fatigue They discount the future or overvalue the present People who have more money than they know what to do with DON'T need to have a really steep discounting functions
27 Next Week Framing, Default Rules and Status Quo Bias Implicit Normativity of Social Science Social Waste 28 Framing of options We seldom consider information in isolated packets (Is this home price of $325,000 a good price?) What is a good option relative to the other available options? (Should I buy this home rather than rent an apartment?) So most options are framed in one way or another Positive and negative framing You have been hoodwinked if you have a preference for one thing over another when the two are equal You don't make these decisions on isolated incidents Framing is an effect that arises because we make decisions 29 Status Quo Bias Leaving things as they are is the status quo. If you follow the default rule, you are preserving the status quo. If you have the systematic tendency to follow the default rule when doing so is far costlier than alternatives, that is the status quo bias. If there is a rule that governs your decision there are default rules that you follow There are default rules for everything There are lots of decisions that we make that is very costly but still we DON'T change strategies The status quo bias has a simple cause and that's very fluent We prefer existing alternatives compared to other alternatives 30 Default Rules Don't check this box and you will/won't be charged the higher rate Take no action, and X will happen Take no action, and this will happen (that is a defualt rule) 31 Acts of Commission Taking an action designed to influence an outcome by that behavior Acts of commission are acts where you do stuff like gross bodily actions If you work longer hours to make more money that is commission If stuff happens to you and you let it happen that is omission Raising the gun and shooting Writing the check to donate
32 Acts of Omission Allowing a foreseeable outcome by not intervening (e.g., not having your child inoculated) Question: Is this "allowing" a decision to allow? Is it a result of inattention, neglect, meanness, cold indifference, etc.? People feel responsible for acts of commissions Inoculation 33 Good and Bad Samaritans Good Samaritans intervene to help another when there is no obligation to do so, and even when doing so carries some costs for the agent Bad Samaritans don't intervene, or take actions to help another, even when doing so is nearly costless to the "agent" 34 Examples Good Samaritan (1) making yourself late for an appointment when you help an old person pick up spilled groceries; (2) you are a bystander who foils a theft at some risk to your safety. Bad Samaritan (1) the last episode of Seinfeld; (2) you are a bystander who does nothing to assist a victim when doing so is at no real cost. 35 What conditions create Acts of Commission and Omission In both cases, the outcome is foreseeable, even if not certain In both cases, a decision may have been made to act or not act If we hold people accountable only for their voluntary bodily movements, then perhaps commission and omission mark an important distinction. If we hold people accountable for their decisions, which are presumably voluntary, then there is no principled difference between commission and omission
36 Default Rules in Organ Donation Moral Default Rules. Between 1995 and 2003, about 45,000 people in the U.S. died waiting for a donor organ. Initially, this is puzzling, since 85% of people in the U.S. support the idea of organ donation. Unfortunately, despite favoring organ donation, less than half have even made a decision about whether to donate, and only 28% have signed a donor card. The U.S. is not unique in enduring this gap between support in principle and action. In Europe, some countries have very high donor rates, and others quite low. 37 What factor(s) determine donation rates? The difference is largely explained by one factor: Whether donation is the country's default consequence of inaction on the part of citizens. By failing to make donation the default option, the government is not just throwing away lifesaving organs, it is letting people die. Because there are untold numbers of supporters who never become donors but would if approached effectively, failure to make a concerted effort to save more lives is a waste of lifesaving parts, throwing good bodies after bad.
38 Default Rules in Insurance Behavioral economics has uncovered yet another case of costly inefficiency, this one quite mundane. In the early 1990s, insurance providers gave drivers in Pennsylvania and New Jersey a choice between two coverage options regarding the right to sue. Drivers in both states were given the choice of a reduced right to sue in exchange for lower insurance rates. But the default option was different for each state. In Pennsylvania, the default option was the full right to sue. That is, if they did nothing, they had the full right to sue. New Jersey drivers had to take action, by selecting the alternate option, in order to acquire this right at an additional cost. 39 The tyranny of default options Now, the two states are demographically similar, so we should expect that the two states have the same percentages selecting the same options. But not so. In fact, the results are startling. Motorists' choices followed the default options in each state. In NJ, where you have to take action for full rights to sue, 20% of the drivers acquired the full right to sue. In PA, where the full right to sue was the default option, 75% of the drivers retained the full right to sue. If New Jerseyans and Pennsylvanians are demographically similar, then either Penssylvanians paid 200 million more than they needed to or New Jerseyans paid 200 million less than they should have (Johnson 2003).
40 The lesson In both the organ donor and the insurance cases, the apparently irrational behavior results in part from a lack of information. We don't already know what organ donation involves or, in the other case, whether the higher insurance premium is worth it. Getting more information takes time, and even then we might not feel confident we made the correct choice. So we follow the default options. But there may be a more efficient means of selecting the best option, one that steers a middle course between simple reliance on default rules and the individual search for all of the available information. 41 Framing of Normative Messages With the number of individuals with self directed pension plans in the millions, this biasharnessing procedure, preventing discounting, has the potential to create a demographic of financially comfortable, rather than desperate, retirees, with all of the benefits that their happiness and welfare can bring. This is an admittedly important financial risk averted, but not all risks are financial. 42 Framing Health Messages Consider `message framing' in the area of health care. Messages carrying the same statistical information can be cast in different ways, to different effect. You can report to the public that people within a certain age range and illness stage who get tested for HIV or breast cancer and test positive have a 70% chance of living beyond 7 years, or a 30% chance of dying in under 7 years. 43 Health Messages Early Detection These messages have different impacts. In order to test the relative effectiveness of health messages, researchers at Yale and the University of Minnesota created videotapes that were either gain or loss framed. The gainframed videos explained the positive effects of healthy behavior and regular breast exams, and the lossframed video attempted to focus attention by frightening the viewer with the bad things that could happen if they don't see a doctor. The subjects in the gainframed message condition were significantly more likely to arrange mammograms. Message framing has been equally successful in motivating HIV testing. It doesn't work very well to frighten people
44 Would a reasonable person be indifferent to the presentation format of a problem if they knew that, were they carrying the illness, the positive frame would make it more likely they would make a decision that would save their life? Once again, the researchers had harnessed a bias the framing bias in order to cultivate behavior that enhances welfare. These are real people in these studies Bias harnessing =balancing off of two biases When you find out that you are a tool you feel kinda cheap and used and people want their preference shaped that way
View Full Document
- Spring '08
- Biology, Elizabeth Loftus, status quo, status quo bias, default rules, Gary Wells