False and true confessions: Are they different?
Have you ever been blamed for something that you didn’t do? Maybe, when you were a child, someone accused you of doing something wrong, like breaking a glass while playing – when it was your brother’s fault instead. However, despite saying that you didn’t do it, the adult kept pressuring you to admit that you were careless and inconsiderate and shouldn’t lie – but you weren’t lying! You keep insisting on your innocence, hoping you will be allowed to continue to do whatever you were doing. But, at some point, in that situation, you realize that the adult has already made up their mind about what happened, and there is only one way to be free: Admitting that you did it. And that’s what you do: “Sorry, I did it. It won’t happen again”. After that, as you expected, you are finally free.
Now, let’s try another scenario and change the characters and context of that situation.
Imagine that instead of breaking something, you are accused of a serious crime, like murdering someone. Instead of an adult, it is a police officer who does not believe you didn’t do it. The frustration you feel for not being believed is multiplied by thousands. The consequence of admitting the blame is not to be freed. But to be locked in prison.
Sadly, you wouldn’t be the first – nor the last – to be in that situation and take responsibility for something you didn’t do. According to the National Registry of Exonerations, as of this writing, 461 people have falsely confessed to a crime they didn’t commit in the United States. These are the lucky ones that have been exonerated. However, others are still in prison, fighting to reopen a case and prove their innocence.
And how does a case get reopened?
“… the odds are not on their side, as it is practically impossible to reopen their cases. But science may still offer hope.”
Either through new evidence or scientific advancements (e.g., a new technique to analyze DNA more accurately). Yet, false confession cases have been characterized as lacking evidence in the first place. In fact, in most cases, false confessions result from coercive police techniques, becoming the primary evidence used to convict them – even in the presence of exculpatory DNA. So, jurors seem to struggle to differentiate between a false and a true confession. Now, you can imagine that the odds are not on their side, as it is practically impossible to reopen their cases. But science may still offer hope.
That’s where my research comes into place.
Police interviews —like any other— result from the interaction between the interviewer and the interviewee, and this dynamic can be captured by dividing the interview into time slots or speaking turns. In 1999, John Pearse and Gisli Gudjonsson analyzed 18 police interviews in 5-minute slots, finding that interviewers used manipulation and coercion to overcome resistance. Interestingly, courts dismissed interviews where apparent pressure was applied. In 2006, Andy Griffiths and Rebecca Milne took a different approach, mapping the dynamic visually by analyzing speaking turns. They plotted question appropriateness on the y-axis and speaking turns on the x-axis—like an electrocardiogram tracking heartbeats. 
And what about false confessions?
The dynamic research in this field is almost nonexistent, but there are grounds to suspect that there are differences between false and true confessions. For instance, false confession statements contained more impersonal pronouns (e.g., “it”) than presumed true statements. Moreover, coercive techniques lead to more false confessions than true ones.
My research
Thus, these previous researchers have established the grounds for the dynamic study of interviews, and I aim to contribute to the field. Specifically, I am comparing false and true confessions to identify whether the interaction is different. For that, I will analyze the content (e.g., the topic), the continent (e.g., how it is done), and the context (e.g., where it is used) of each speaking turn in both interviewer and interviewee. Then, with the codebook, I will train a Large Language Model to classify all these elements. This way, I will reduce any possible inaccuracy, and future researchers can apply and improve this codification.
Once I have the first layer of codification (i.e., speaking turns), I will identify recurring patterns that frequently happen throughout the interviews: these would be the techniques. This will allow me to describe the interviews using techniques and explore if they are differently structured.
The final step is to map speaking turns and techniques together, similar to previous researchers, and provide a complete “electrocardiogram” tracking speaking turns.
Ultimately, these linguistic patterns suggest that how a confession unfolds could be the key to distinguishing between true and false confessions. Advancing the science of police interviewing may not only help correct past injustices but also prevent new ones.