When a friend offers to set up a blind date, the question is always how much you trust this person’s judgment. We all know people who equate loud with friendly and sarcastic with smart, so it can be hard to accept other people’s impressions.

But would you accept a computer’s?

Linguistics researchers at places like Columbia and Stanford are developing computer programs that may be better at reading people than even other people. Dr. Julia Hirschberg of Columbia is programming computers to break down speech to find signs of lying, while a professor at the University of Southern California is teaching them to recognize anger.

Ok, so the computers will probably never be able to help men understand women and vice versa, but they still could help clear up some of those interactions too. Dr. Dan Jurafsky, of Stanford has been analyzing the language used by speed daters. He’s working on identifying qualities like friendliness and flirtatiousness.

Dr. Jurafsky says that the goal is to build systems that can understand emotions from speech. But some people are suggesting that this type of research may have practical uses. For example, an online dating site could recommend a potential match as friendly or not, based on recorded sample conversations.

In an ideal world, maybe this would work and even be helpful. But I don’t think that most of us would be able to trust it. The same friend who thinks loud equals friendly might think “This person’s not friendly! Just look at how quiet he/she is!” and give up on trusting the computer for good.

We all have some strange definition for a quality that we admire and look for in a date and most of us probably have no idea what it is. What I see as defining friendliness, kindness, intelligence or humor may not be how the rest of the world, or a computer, define them.