Imagine a poor, strung-out woman who testifies and then a man in a suit gives contradictory testimony. A judge hears both parties and rejects the woman's testimony.
Our judicial system depends on having people testify in court. The theory is that judges and juries can tell if someone is lying by looking at them while the person speaks.
The problem with the theory underpinning our oral testimony scheme is that judges can't tell who's lying. It turns out that decades of psychological research has shown that humans are very bad at detecting lies (~54% of the time people can detect a lie, Vrig et. al, 2010). I've only provided one citation for that bold claim but if you Google around you'll come up with hundreds of other papers on the subject, layman articles that summarize the papers and essays in The New Yorker.
We're applying an ancient theory about truth at a time when we know that the method doesn't work.
Who pays for this unscientific approach to the truth? From a 2010 Legal and Criminological Psychology paper,
"... using a mock jury paradigm, Porter, Gustaw, and ten Brinke (2009) presented participants with vignettes of the same crimes accompanied either by a photo of a defendant whose face previously had been rated (in a pilot study) as appearing highly trustworthy or untrustworthy. It was found that participants required less evidence before finding an untrustworthy defendant guilty of murder. When presented with major exonerating evidence (DNA implicating someone else), participants changed their verdict to not guilty 84% of the time for trustworthy-looking defendants but only 42% of the time for untrustworthy-looking ones. Previous work had already established that baby-faced individuals receive more lenient judicial outcomes than mature-faced individuals and African American offenders with more Afrocentric facial features receive harsher sentences than those with fewer such features (Blair, Judd, & Chapleau, 2004; Zebrowitz & McDonald, 1991). Further, attractive defendants are more likely to be found not guilty, handed shorter sentences, and considered less dangerous than unattractive individuals (e.g. Bull & Rumsey, 1988; Downs & Lyons, 1991; Esses & Webster, 1988)."
In my example at the beginning of this post the judge was likely applying their own biases about previous drug-using women who've testified while also judging the information they've been presented with. They've spoken to many men in suits who've told the truth (or at least they think so) and heard from many drug users who lied. These circumstances can't help but affect how someone's honesty is judged. And the worst part about this situation is that individuals are wildly overconfident in their ability to judge people (such as during employment interviews, now considered a very poor way of judging candidates by social scientists but still popular with employers).
If our legal system used written evidence instead of oral testimony we could eliminate the biases and pre-judgement of what people have to say while probably losing nothing in accuracy (since people are so bad at telling the truth anyway). We'd end up with a system that's a lot more fair to the people whom society doesn't treat all that well. As a bonus, this system would be much cheaper and easier to organize, perhaps taking place entirely online.