9.10: Overconfidence

Problem 1 (adapted from Alpert & Raiffa, 1969):

Listed below are 10 uncertain quantities. Do not look up any information on these items. For each, write down your best estimate of the quantity. Next, put a lower and upper bound around your estimate, such that you are 98 percent confident that your range surrounds the actual quantity. Respond to each of these items even if you admit to knowing very little about these quantities.

1. The first year the Nobel Peace Prize was awarded

2. The date the French celebrate "Bastille Day"

3. The distance from the Earth to the Moon

4. The height of the Leaning Tower of Pisa

5. Number of students attending Oxford University (as of 2014)

6. Number of people who have traveled to space (as of 2013)

7. 2012-2013 annual budget for the University of Pennsylvania

8. Average life expectancy in Bangladesh (as of 2012)

9. World record for pull-ups in a 24-hour period

10. Number of colleges and universities in the Boston metropolitan area

On the first problem, if you set your ranges so that you were justifiably 98 percent confident, you should expect that approximately 9.8, or nine to 10, of your ranges would include the actual value. So, let’s look at the correct answers:

Figure 7. Overconfidence is a natural part of most people's decision-making process and this can get us into trouble. Is it possible to overcome our faulty thinking? Perhaps. See the "Fixing Our Decisions" section below. [Image: Barn Images, goo.gl/BRvSA7]

1. 1901

2. 14th of July

3. 384,403 km (238,857 mi)

4. 56.67 m (183 ft)

5. 22,384 (as of 2014)

6. 536 people (as of 2013)

7. $6.007 billion

8. 70.3 years (as of 2012)

9. 4,321

10. 52

Count the number of your 98% ranges that actually surrounded the true quantities. If you surrounded nine to 10, you were appropriately confident in your judgments. But most readers surround only between three (30%) and seven (70%) of the correct answers, despite claiming 98% confidence that each range would surround the true value. As this problem shows, humans tend to be overconfident in their judgments.

In 1984, Jennifer Thompson was raped. During the attack, she studied the attacker's face, determined to identify him if she survived the attack. When presented with a photo lineup, she identified Cotton as her attacker. Twice, she testified against him, even after seeing Bobby Poole, the man who boasted to fellow inmates that he had committed the crimes for which Cotton was convicted. After Cotton's serving 10.5 years of his sentence, DNA testing conclusively proved that Poole was indeed the rapist.

Thompson has since become a critic of the reliability of eyewitness testimony. She was remorseful after learning that Cotton was an innocent man who was sent to prison. Upon release, Cotton was awarded $110,000 compensation from the state of North Carolina. Cotton and Thompson have reconciled to become close friends, and tour in support of eyewitness testimony reform.

One of the most remarkable aspects of Jennifer Thompson’s mistaken identity of Ronald Cotton was her certainty. But research reveals a pervasive cognitive bias toward overconfidence, which is the tendency for people to be too certain about their ability to accurately remember events and to make judgments. David Dunning and his colleagues (Dunning, Griffin, Milojkovic, & Ross, 1990) asked college students to predict how another student would react in various situations. Some participants made predictions about a fellow student whom they had just met and interviewed, and others made predictions about their roommates whom they knew very well. In both cases, participants reported their confidence in each prediction, and accuracy was determined by the responses of the people themselves. The results were clear: Regardless of whether they judged a stranger or a roommate, the participants consistently overestimated the accuracy of their own predictions.

Eyewitnesses to crimes are also frequently overconfident in their memories, and there is only a small correlation between how accurate and how confident an eyewitness is. The witness who claims to be absolutely certain about his or her identification (e.g., Jennifer Thompson) is not much more likely to be accurate than one who appears much less sure, making it almost impossible to determine whether a particular witness is accurate or not (Wells & Olson, 2003).

I am sure that you have a clear memory of when you first heard about the 9/11 attacks in 2001, and perhaps also when you heard that Princess Diana was killed in 1997 or when the verdict of the O. J. Simpson trial was announced in 1995. This type of memory, which we experience along with a great deal of emotion, is known as a flashbulb memory—a vivid and emotional memory of an unusual event that people believe they remember very well. (Brown & Kulik, 1977).

People are very certain of their memories of these important events, and frequently overconfident. Talarico and Rubin (2003) tested the accuracy of flashbulb memories by asking students to write down their memory of how they had heard the news about either the September 11, 2001, terrorist attacks or about an everyday event that had occurred to them during the same time frame. These recordings were made on September 12, 2001. Then the participants were asked again, either 1, 6, or 32 weeks later, to recall their memories. The participants became less accurate in their recollections of both the emotional event and the everyday events over time. But the participants’ confidence in the accuracy of their memory of learning about the attacks did not decline over time. After 32 weeks the participants were overconfident; they were much more certain about the accuracy of their flashbulb memories than they should have been. Schmolck, Buffalo, and Squire (2000) found similar distortions in memories of news about the verdict in the O. J. Simpson trial.