‘Lucy is a gifted communicator, makes occasional gut decisions against logic and prefers big ideas over details.”

As a description of me, this is pretty good. I have spent my working life communicating, so I jolly well ought to be good at it by now. I have just made a decision in my private life which I know is not only illogical but financially catastrophic, yet I’m doggedly going with my gut. And as for details, unless they suit my argument, I have no fondness for them at all.

Yet what is so disturbing about the above character analysis is that it does not come from someone who knows me; it was thrown together in three seconds flat by an algorithm developed by some computer scientists who have worked out how to trawl the internet for all public information about a person and turn it into a potted portrait of a personality.

Having let Crystalknows. com analyse me, I then typed in my colleagues' names and found they were summed up uncannily well too. A fellow columnist, whom I know to be impatient, bold and creative, was judged to be just that by Crystal. Some of its verdicts were a bit off the mark. Yet in each case, the app tells you how confident it is; the cases in which accuracy was poor were mostly those where there wasn't much data to go on.

Suggested approach

This is all good fun, but the main selling point of Crystal is not to provide uproarious entertainment. It is to help you communicate better. You simply link the app to Gmail and every time you start writing a message it tells you what sort of approach will go down best with the recipient.

I have just emailed Drew D'Agostino, the founder of Crystal Projects. "Be brief", warned a little green button at the bottom of my screen, telling me that Drew doesn't like wordy. Had I been writing to someone else, it might have advised: be casual or be formal.

According to D’Agostino, Crystal is the biggest improvement to email since the spell check. But so far tetchy bloggers are not impressed. It’s been called the “stalking app” and “creepy” and “sinister”. Some people have protested that it doesn’t feel good to know perfect strangers are looking you up and forming instant views of you.

Yet none of this strikes me as particularly creepy – all the data is already in the public domain. And I think I’d rather that decisions about me were made based on some sort of system, rather than on hunch, prejudice and ignorance.

Unlike most apps, Crystal solves a genuine problem. There are no common rules on email. Every time we write an email we are in the dark as to what is going to work best.

You could say that a post-Crystal world, in which all messages arriving in my inbox were written in the same prescribed style, would be a dull one. Yet it couldn’t be as dull as the current arrangement whereby I read dozens of emails every day that begin with windy small talk like “I hope this email finds you well”. If people knew how much I hated that, they would desist. That would save time for them and save me from irritation. Win: win.

A more philosophical complaint is that if all the judging of character were done by machines, we would lose our own – more nuanced – ways of assessing people. And because we are suggestible, we don’t question the apps’ results, even when they are based on not much information. Confirmation bias and all that.

Basic objection

There may be something in this, but my main objection to Crystal is more basic. It doesn’t work well enough. I should be the easiest person in the world for Crystal to analyse given that I have published a personal column every week for 21 years. Even though the app has more or less got my character right, its suggestions on how to approach me are worse than useless.

“Use emoticons,” it begins. Is this a joke? I have never knowingly used an emoticon in my life, and instantly knock a couple of points off the IQ of anyone who deploys them. Even more outrageously, it says that I don’t mind if people are late. Again: could not be more wrong. I am obsessively punctual.

Emoticonphobia and obsessive timekeeping are not things that an algorithm needs to deduce. They are phobias that I have written about explicitly and often. Thus my problem with Crystal is not that it is stalking me, but that it doesn’t stalk nearly enthusiastically enough.

In my brief email to Drew, I asked about accuracy. “Hi Lucy. Thanks for reaching out,” he emailed back, thus committing two email bloopers before he had even got going. He then reassured me that accuracy is getting better, and soon users will be able to correct mistakes and supply preferences themselves.

In the meantime, here is my character report on Crystal: great idea; must try harder. – Copyright The Financial Times Limited 2015