Feelings or algorithms: how we make choices in the digital age

How digital tools filter everyday choices, why people retain personal preferences, and where random selection removes habits that offer no mathematical advantage.

There are things that happen on a daily basis that evolve around the tech we use that we don’t even know about. The email app decides what looks like spam. The phone groups photographs by face and location. An online shop rearranges its front page depending on what was searched for last week. By the end of the day, dozens of small choices have already been filtered before a person gets anywhere near them. Usually that is welcome. There are too many emails, too many products and too many things competing for attention for anyone to inspect all of them properly. Software cuts down the pile. What is interesting is that people still become very particular when the choice feels personal, even when the thing being chosen is completely arbitrary.

Numbers bring out some strange habits

Give somebody a list of eighty numbers and ask them to pick ten. A supposedly random choice can become surprisingly personal. Birthdays appear. So does the number on an old football shirt, a house number or the date of an anniversary. Some people avoid consecutive numbers because they look unlikely. Others spread their choices across the page because a cluster somehow feels wrong. None of that makes a selection more random. This is one of the few cases where handing the job to software can actually make the choice less complicated. A keno number generator simply produces a set of numbers without caring whether 7 has always been somebody’s favourite or whether 41 and 42 look too close together. It is not making a prediction. It is just taking personal preference out of a task where personal preference has no mathematical advantage. The same basic idea appears in much less obvious places. Offices use random draws when assigning names. Researchers rely on randomisation when selecting samples. Software testers generate large sets of random inputs because a human tester will naturally repeat familiar patterns. People are very good at finding meaning in numbers, which is useful in plenty of situations and completely irrelevant in others.

We still keep the last word

Things become less straightforward when the computer is not choosing randomly but trying to guess what we want. A ticket site may suggest the “best” seat based on price and view, yet somebody will still move two rows back because they prefer being near an aisle. A photo app can identify hundreds of technically good pictures from a holiday, while the one that ends up framed may be slightly blurred because of who happens to be in it. Even grocery apps can learn what a household buys every week and still fail the moment somebody decides they are bored of the usual dinner. This is the part that software has a hard time with. Data is useful when yesterday is a decent guide to tomorrow. People are not always that consistent. The algorithms behind digital services are getting better at sorting, ranking and removing choices that are probably irrelevant. That saves time, but it does not necessarily make the last decision any more logical. A person can use an app to compare twenty options in seconds and then choose the one that simply feels right.

So what do we have

Perhaps that is why the arrangement works as well as it does. Software handles the volume; people keep the odd preferences, old habits and sudden changes of mind that do not fit neatly into a recommendation model. And when none of those preferences should matter, sometimes it is easier to let the machine pick the numbers.