In this experiment we set out to examine how people process potentially irrelevant information when they already established certain pre-existing beliefs. To uncover the mechanism behind confirmation bias, we ask subjects to report beliefs of the assigned urn, in which they have prior beliefs and a piece of potentially irrelevant information. Crucially, they also have to report beliefs of the irrelevant urn, by which we can visually observe the strength of weight they put on the potentially irrelevant information. We show that subjects tend to view this information as com-pletely worthless in evaluating the assigned urn when it conflicts their prior beliefs, but overvalue it when it confirms their prior beliefs. We estimate the tendency of attributing the information to the irrelevant urn. The results suggest that on av-erage subjects believe the information is from the irrelevant urn with probabilities more regardless of the types of information. However, they increase the probabilities when the information is conflicting by 12%. When we allow subjects consider other’s information might be inaccurate, the they still believe the information is more likely from the irrelevant urn when it is conflicting. These results are robust even we assume subjects independently make decisions on the assigned and irrelevant urn.
Most importantly, we try to explore the mechanism leading to the echo chamber, especially focusing on the information updating. By explicitly creating an irrelevant urn, we highlight one possible reason people usually stick to their political stance or beliefs on controversial issues, even leading to polarization. Though this may not be the only cause of the echo chamber e↵ect, our results suggest that dismissing the information when it conflicts with one’s prior is still a prominent cause.
References
L. Babcock, G. Loewenstein, S. Issacharo↵, and C. Camerer. Biased judgments of fairness in bargaining. American Economic Review, 85(5):1337–1343, 1995.
M. A. Costa-Gomes and V. P. Crawford. Cognition and behavior in two-person guessing games: An experimental study. American Economic Review, 96(5):1737–
1768, 2006.
A. Coutts. Good news and bad news are still news: Experimental evidence on belief updating. Experimental Economics, 22(2):369–395, 2019.
D. Eil and J. M. Rao. The good news-bad news e↵ect: asymmetric processing of ob-jective information about yourself. American Economic Journal: Microeconomics, 3(2):114–38, 2011.
U. Fischbacher. z-tree: Zurich toolbox for ready-made economic experiments. Ex-perimental Economics, 10(2):171–178, 2007.
R. G. Fryer Jr, P. Harms, and M. O. Jackson. Updating beliefs when evidence is open to interpretation: Implications for bias and polarization. Journal of the European Economic Association, 17(5):1470–1501, 2019.
E. L. Glaeser and C. R. Sunstein. Why does balanced news produce unbalanced views? Technical report, National Bureau of Economic Research, 2013.
B. Greiner. Subject pool recruitment procedures: organizing experiments with orsee.
Journal of the Economic Science Association, 1(1):114–125, 2015.
D. M. Grether. Bayes rule as a descriptive model: The representativeness heuristic.
The Quarterly Journal of Economics, 95(3):537–557, 1980.
C. A. Holt and A. M. Smith. An update on bayesian updating. Journal of Economic Behavior & Organization, 69(2):125–134, 2009.
C. A. Holt and A. M. Smith. Belief elicitation with a synchronized lottery choice menu that is invariant to risk attitudes. American Economic Journal: Microeco-nomics, 8(1):110–39, 2016.
D. M. Kahan, H. Jenkins-Smith, and D. Braman. Cultural cognition of scientific consensus. Journal of Risk Research, 14(2):147–174, 2011.
D. M. Kahan, E. Peters, M. Wittlin, P. Slovic, L. L. Ouellette, D. Braman, and G. Mandel. The polarizing impact of science literacy and numeracy on perceived climate change risks. Nature Climate Change, 2(10):732–735, 2012.
C. G. Lord, L. Ross, and M. R. Lepper. Biased assimilation and attitude polariza-tion: The e↵ects of prior theories on subsequently considered evidence. Journal of Personality and Social Psychology, 37(11):2098, 1979.
A. Tversky and D. Kahneman. Availability: A heuristic for judging frequency and probability. Cognitive Psychology, 5(2):207–232, 1973.
Appendix
A First Phase Belief
The data points aligned with 45 degree line in the irrelevant urn, implying that subjects believe the initial draw can infer both urns. Figure9a shows that a majority of these choices are made by di↵erent subjects and they only perform this behavior one time. Moreover, Figure 9b shows the occurred round of these choices. They do not concentrate on particular rounds, suggesting that such unusual behavior is randomly made throughout the experiment and is unlikely explained by learning e↵ect.
Figure 9: Beliefs Aligned with 45 Degree Line in the Irrelevant Urn. (a) the Number of Rounds (b) Occurrence Rounds.
B Second Phase Raw Data
Figure10shows the raw data of second phase beliefs. In particular, it is clear to see the overreaction in the irrelevant urn.
Figure 10: Elicited Beliefs in the Second Phase of the (a) Assigned (b) Irrelevant Urn
C Alternative Experimental Designs
We document alternative designs that were eventually dropped. Our first experi-mental design is inspired byEil and Rao (2011). Subjects are asked to predict the real value of an asset with ten possible states. The computer randomly draws with replacement three balls from twelve, in which ten balls represent the ten possible states and the additional two balls represent the real value. Thus, the real value is drawn with probability 0.25 compared to others with 0.083. After observing their private information of three ball draws, they report their beliefs of each state that add up to 1.
Subjects then observe new information: The computer divides others into two halves, one half whose predictions are close to and the other half whose predictions are far from the subject, and randomly draws another subject from one of them to reveal his/her prediction. The procedure is repeated three times, so three other subjects’ predictions will be revealed to the subject. We elicit beliefs in terms of
probabilities after subjects observe each piece of information using the quadratic scoring rule. The experimental interface is shown in Figure11.
Figure 11: Screen Shot of the First Version Experiment.
Our second experimental design is similar to the first one, but with only two possible states. There are two urns, A and B, in the experiment. Urn A applies the Maximum Rule and Urn B applies the Minimum Rule, so each urn reports either maximum or minimum of two draws from the uniform distribution. We provide the probability table in case subjects cannot figure it out themselves. Subjects observe a ball from urn A or B with equal chance, and report the probability that the chosen urn is A. Then, subjects observe others’ information and beliefs are elicited using the same design as the first version.
Our third experimental design is nearly identical to our final one implemented, but with three important di↵erences. First of all, it is a one shot game with three stages of belief-updating, while the final experiment has ten rounds each with one
stage of belief-updating. In other words, subjects observe their initial draw and then receive three other piece of information. Second, we use the BDM procedure as inCoutts (2019) to elicit beliefs, which is illustrated in Figure 12a. Finally, the probability for drawing each number under the Maximum Rule and Minimum Rule is shown in tables. The experimental interface is shown in Figure 12b.
Figure 12: Screen Shot of Third Version Experiment.
TASSEL p.1
TASSEL p.2
TASSEL p.3
35
1
1. 35 35 36 100
1
1
2. 35 34 33 0
1 2018 12
18 C 1
2018 12
18 C
2018 12 18
C 2018 12
18 C
TASSEL p.4
TASSEL p.5
TASSEL p.6
1.
2.
6
1 1