A couple of years ago, my mum asked me to take a look at her new hearing aids. Despite being more powerful (and expensive) than her last pair by a large margin, she simply could not understand how it decided which sounds to amplify and which were “background noise” to filter out, leaving her with the constant feeling she might be missing something.

A quick look at the settings made it obvious that on top of standard audio processing, the hearing aids were likely using a machine learning algorithm that had been trained to recognise and suppress sounds with certain characteristics as “noise”. The telltale sign was the sheer opacity of the controls: a handful of sliders with vague, subjective names like “background chatter” that never seemed to do quite what they implied. In a nutshell, the designers had made a more powerful device by sacrificing user understanding and control.

This is very much a programmer’s idea of user-centred design, and it is one I am all too familiar with from working in the medical and assistive technology spaces. Any problem can be solved with a better algorithm, and better means smarter. This does not mean complex – I doubt many of us could build a watch or a scientific calculator, and yet you would be hard pressed to call either a smart device. Instead, I would argue the essence of “smartness” in tech is moving decision-making power from the user to the machine itself. After all, if the machine can perfectly anticipate what the user wants, why would they need control?

The obvious problem with this is that no algorithm can perfectly anticipate what any possible user might want but, speaking cynically, this shortcoming is very good for business. The pursuit of the perfect system provides an excuse for endless tweaks by highly-paid software engineers, endless new versions with new features that users must pay for, then pay for the removal of once they prove annoying and unpopular. None of this is necessarily malicious, but if you concentrate a large number of (not very diverse) techno-optimists in one place and then make them the model for modern industry, it is not very surprising either.

“Dumb” machines are therefore a species in decline. After years of feature creep, simplicity acquires the air of stagnation, and a bare-bones begins to look like terrible value for money, even if you don’t use 70% of the features you’re paying for right now. It is also tremendously embarrassing to admit you don’t get the new big thing – in our current society, and particularly in tech, this is the signal that you have officially become “old”, soon to be left behind. However the purpose of technology is not to be new or exciting, it is to do things, and we lose sight of that at our peril.

I still love designing strange new algorithms (my PhD thesis is an exercise in cybernetic esoterica), but I have a lot of affection for dumb machines. You don’t have to turn off your phone’s autocomplete in an over-optimistic attempt to triage your dreadful spelling as I have (to mixed success), but I do encourage you to think critically about the decisions their devices and software make for you, and whether you like them. Because if you don’t, there are always other options.