Five-letter words for research: ‘trial’ and ‘error’

How brilliant would it be if research worked in the same way [as Wordle]!

Dr Matthew Partridge

At the start of the year, I found - like so many others - that my social media feed had been infected by a strange bug where all the posts were titled something like "219 3/6" and were made up of strange grey, yellow, and green squares. I am now aware that this was my first sighting of the global word game craze Wordle. Released towards the latter end of 2021, Wordle took the internet by storm in early 2022 when the creator added the share feature that quickly suffused my social feeds with the small, coloured squares neatly laid out in grids of five across.

For those of you that haven't heard of this craze; Wordle is an online word game where each day you have to guess a mystery five-letter word. You get six guesses and each time you guess you are told if any of the letters are correct (yellow letters are right but in the wrong place, green letters are right and in the right place) so that you can narrow down your next guess. Each guess gets you that bit closer to the right answer and helps you slowly come to the only answer it can be.

How brilliant would it be if research worked in the same way!

You get your results back from your experimental first run. All black except a few points in one corner which are yellow. You repeat your experiment, this time the points in the corner have moved along and are now green with some more yellows either side. You repeat the experiment a third time and now everything is black again because you actually used the wrong concentration, but you get the idea. I mean to some extent this does already happen. I've worked for supervisors before that would happily return my results in colour coded format, with the bits they think are wrong and with a note to come back when I get them all right. But unlike Wordle, that colour coding was only red and more based on wishful thinking than accuracy.

What if we had a more accurate version. What if we worked on building computers that didn't necessarily know the exact answer but knew the likely answers as well as the answers it couldn't be. So when we run experiments we can check them against the computer simulations and...wait...I think I just invented machine learning... but with colourful boxes.

Admittedly adding a nice user interface and some colourful boxes would do the field of machine learning a world of good but I think we're still some way off being able to 'Wordle' research with it. Which is a shame because it sounds like a fun way of working. Maybe the best thing to do is not worry about making research better but to make learning science more fun. I would have learnt the Krebs cycle a lot quicker if I'd had a Wordle-esk interface where I had to guess the correct order of intermediates. Or, considering how Wordle has managed to teach millions all about common letter combinations, I would think it could do wonders for teaching organic molecule naming conventions.

Now, sorry to cut this article short but I'm stuck on today's Wordle and I've already had five tries and only got two green letters and one of them was a lucky typo…

Author: Dr Matthew Partridge is a researcher, cartoonist and writer who runs the outreach blog  errantscience.com