The ALT key: how innocent people come to confess

The student sat at a keyboard, typing letters as another student read them aloud. Before starting, they had been warned: do not press the ALT key, or the computer will crash and the data will be lost. Partway through, the computer stopped. The experimenter accused the student of pressing ALT. Every student in the study was innocent; the crash was rigged. Every one of them denied it at first (Kassin & Kiechel, 1996).

Saul Kassin and Katherine Kiechel then changed two things. For some students the letters came quickly, so they could not be sure what their fingers had done. For some, the person reading the letters, who was working for the researchers, said she had seen them hit the key. Then the experimenter wrote out a confession and asked the student to sign it (Kassin & Kiechel, 1996).

Students who were rushed and had a "witness" against them were the most likely to sign. They were also the most likely to believe it: to tell someone afterwards that they had hit the wrong key, and to supply details of how they did it. Nothing had happened. They remembered it anyway (Kassin & Kiechel, 1996).

The stakes were a broken computer. In an interview room they are years of someone's life.

Three kinds of false confession

The study's title names what it found: compliance, internalisation and confabulation (Kassin & Kiechel, 1996). They map onto three ways an innocent person ends up saying "I did it".

• Compliance. The person knows they did not do it, and says they did to make the pressure stop. The confession is a trade: an end to the interrogation now, for a cost that feels far away.
• Internalisation. The person comes to believe they did it. Uncertainty and a confident accuser are enough.
• Confabulation. The belief grows details. The person "remembers" how it happened, and those details make the confession sound true.

A team of researchers who study confessions, Kassin, Steven Drizin, Thomas Grisso, Gisli Gudjonsson, Richard Leo and Allison Redlich, gathered what is known about when this happens in real cases. In the person: youth, intellectual disability and mental illness. In the room: long interrogations, police presenting evidence that does not exist, and minimisation, the quiet suggestion that confessing will go easier (Kassin et al., 2010).

The deal that makes truth harder to find

Melissa Russano and colleagues built a second experiment to test two of those tactics on guilty and innocent people side by side. Students worked on problems in pairs. Some were drawn into helping their partner, which broke the study's rule; others were not. Then everyone was accused of cheating (Russano et al., 2005).

The guilty confessed more than the innocent, as they should. But minimisation and an explicit offer of a deal each raised confessions from both groups. The tactics pulled out more true confessions and more false ones at the same time (Russano et al., 2005). A confession produced that way tells an investigator less than one produced without it. The authors advised police to avoid any technique that implies or promises leniency, because it lowers the value of the confession it obtains (Russano et al., 2005).

Nobody can tell by watching

A false confession would matter less if experienced people could spot one. Kassin, Christian Meissner and Rebecca Norwick tested that. Ten prison inmates were recorded confessing to crimes. Five confessions were true. Five were made up for the study. College students and police investigators watched or listened and judged each one (Kassin et al., 2005).

The students were more accurate than the police. The investigators were more confident, and they leaned towards calling confessors guilty. When the researchers told a second group that exactly half the confessions were false, the investigators' lean towards guilt disappeared, but their accuracy did not improve and their confidence did not drop (Kassin et al., 2005).

Once a confession exists, it travels. Forensic examiners who know a suspect has confessed read ambiguous evidence differently, and so do witnesses (Kassin et al., 2013). One manufactured statement can recruit the rest of the case to agree with it.

Two ways to question a suspect

Meissner, Redlich, Sujeeta Bhatt and Susan Brandon reviewed the research for the Campbell Collaboration. They describe two broad methods. The accusatorial method, used mainly in the United States and Canada, is built on accusation, confrontation, psychological manipulation and refusing to accept denials. The information-gathering method, used in the United Kingdom, New Zealand, Australia and much of Western Europe, is built on rapport, seeking the truth and active listening (Meissner et al., 2012).

In the field studies, both methods produced confessions at similar rates. In the experiments, where the researchers knew who was guilty, information-gathering raised true confessions and lowered false ones (Meissner et al., 2012). The review calls the finding preliminary, because there were only five field studies and twelve experiments to pool (Meissner et al., 2012).

What protects the innocent

The fixes come from the same research, and none is expensive (Kassin et al., 2010; Meissner et al., 2012):

• Record every interrogation in full, from the first question, so a court can see how a confession was produced, not only what it says.
• Do not tell suspects there is evidence against them that does not exist.
• Do not imply leniency in exchange for a confession (Russano et al., 2005).
• Give young people, and people with intellectual disability or mental illness, protection in the room: an adult or lawyer who is on their side.
• Ask open questions and listen, rather than accuse and refuse denials.
• Keep the confession away from forensic examiners until their own work is written down (Kassin et al., 2013).

Each one assumes something uncomfortable: that a confession is evidence like any other, made by people under pressure, and that the people applying the pressure cannot see its effect. Without them, a justice system can produce guilt built at scale one signature at a time.

The students in 1996 never touched the ALT key. Some of them could tell you how they did it.

Short answers

Why would an innocent person confess?

Research identifies pressures in the interrogation and traits in the person. In the room: long questioning, being told there is evidence against you that does not exist, and minimisation, which makes confessing sound like the way out. In the person: youth, intellectual disability and mental illness (Kassin et al., 2010).

Can police tell a true confession from a false one?

Not reliably. When college students and police investigators judged ten recorded confessions, half of them false, students were more accurate than police; investigators were more confident and leaned towards judging confessors guilty (Kassin et al., 2005).

Which interview method produces fewer false confessions?

A systematic review found that in experiments with known ground truth, information-gathering interviewing (rapport, open questions, listening) raised true confessions and lowered false ones compared with accusatorial interrogation. The authors call the findings preliminary because the studies are few (Meissner et al., 2012).

References

1. Kassin, S. M., & Kiechel, K. L. (1996). The social psychology of false confessions: Compliance, internalization, and confabulation. Psychological Science, 7(3), 125–128. doi.org/10.1111/j.1467-9280.1996.tb00344.x
2. Kassin, S. M., Drizin, S. A., Grisso, T., Gudjonsson, G. H., Leo, R. A., & Redlich, A. D. (2010). Police-induced confessions: Risk factors and recommendations. Law and Human Behavior, 34(1), 3–38. doi.org/10.1007/s10979-009-9188-6
3. Kassin, S. M., Dror, I. E., & Kukucka, J. (2013). The forensic confirmation bias: Problems, perspectives, and proposed solutions. Journal of Applied Research in Memory and Cognition, 2(1), 42–52. doi.org/10.1016/j.jarmac.2013.01.001
4. Kassin, S. M., Meissner, C. A., & Norwick, R. J. (2005). “I’d know a false confession if I saw one”: A comparative study of college students and police investigators. Law and Human Behavior, 29(2), 211–227. doi.org/10.1007/s10979-005-2416-9
5. Meissner, C. A., Redlich, A. D., Bhatt, S., & Brandon, S. (2012). Interview and interrogation methods and their effects on true and false confessions. Campbell Systematic Reviews, 8(1). doi.org/10.4073/csr.2012.13
6. Russano, M. B., Meissner, C. A., Narchet, F. M., & Kassin, S. M. (2005). Investigating true and false confessions within a novel experimental paradigm. Psychological Science, 16(6), 481–486. doi.org/10.1111/j.0956-7976.2005.01560.x