What teaching 10 year olds rugby has taught me about AI
There’s a lot of interest at the moment on how Artificial Intelligence can create the “perfect” algorithm, or whatever it might be. But whilst that can be very valuable for particular tasks, it’s important not to lose the human or cultural factors when looking at how your organisation uses data.
I have recently been “roped in” to being an additional coach for my son’s rugby team. The boys, and girls, are only 10, so are still learning – they have only just started doing uncontested three-man scrums, if you are interested – but the coaches asked for extra support, so I volunteered. As part of that I attended the “Rugby Ready” course, which is the basic training course requirement from the Rugby Football Union.
This was particularly interesting in what it taught about the ethos and spirit of the game. One of my main personal reasons for doing the coaching was to understand the changes to the rugby laws since I last played 20 years ago. The laws for the different school age groups also change as you go up, and there is much more about player safety than there was when I played, which was when the game was still amateur at all levels. All this means that most of the time on the side-lines I have been baffled by what the referees in these junior matches have been deciding. Given Station10’s business background, I am a very analytical person, so want to know exactly what is going on. If I was honest with myself, I always had the idea in the back of my mind that a rugby match can be refereed perfectly, where every infringement is spotted, clearly identified and communicated fairly, but still allowing for a free-flowing game of rugby (and ideally with our team winning!). A bit like some sort of all-seeing, all-knowing AI system, and, in my mind, anything that didn’t achieve that was a failure. Until learning to coach kids about rugby, I never realised the ridiculous internal contradiction that such unfeasible expectations contained.
My more recent experience of rugby before my son started playing has been watching televised and occasionally live games, mostly internationals and Premiership level. Of course, at these top levels, massive investment and technological progress have supported referees with TV cameras and replays, allowing for infringements, errors, and moments of try-scoring brilliance to be shown again in slow motion for match officials to ponder and reconsider during the match. The use of this in-game technology by referees is what differentiates rugby from football in particular.
But it has had perhaps the unfortunate side-effect that this idea of a perfectly managed game is achievable. The top levels of the now-professional game also attract the very best referees, who have the knowledge, experience, personal authority and communication skills to police a game that is getting more physical and combative. It is perhaps a credit to those top-level refs that the idea of a perfect AI-officiated game seems at all possible.
On reflection, however, it’s not the all-seeing eye that is the most important component – it’s the communication skills. Players respect the refs the most if they communicate what they see, and can adjust their game accordingly; players realise that refs are only human and are doing the best that they can. The communication skills are what stand out – there are entire videos of Nigel Owens, the best current rugby referee in the world, explaining his decisions and enforcing player discipline in a very engaging, witty and often humorous way. When you think about it, this humour and ethos is the least “artificially intelligent” part of the perfect game.
It’s this part of the sport that I have been reminded about in my training course. The trainer, a well-built, early-middle-aged, bearded veteran of many a ruck and maul, bemoaned the fact that the atmosphere at Twickenham the previous weekend resembled more of a football crowd than a rugby one, with types of language from the stands much more aggressive and frequently inappropriate than the spirit of rugby outlines. Because there is a spirit of rugby. Unusually, the game does have a set of core values, summarised by the acronym TREDS, to which those involved should adhere. The R, for example, stands for Respect; of players, supporters, and officials. One thing that does remain sacrosanct in rugby is the referee’s decision is final, and is not argued with on the field of play. I hadn’t realised, but if any player pushes or attacks a rugby referee, they will be served with an effective life ban from the game (compare that to football, either association or American). Players, and indeed referees, are not automatons; they are humans, and should be treated accordingly.
This is particularly true with two teams of 10-year-olds; in this scenario, the referee (almost always someone’s dad at the club) is just doing their best, but the most important focus, apart from player safety, is letting the kids enjoy themselves. The best referees are those who explain and communicate to the children what is going on, and what they can and can’t do. At this level, they let certain minor infringements go in the name of enjoyment. But they also need to communicate with the supporters. The difficulty can be that you have a group of other dads on the touchline who are watching to see how well little Jonny does, and are likely to be much more attuned to, and upset by, any incident when he might be hurt or hard done by. And, fed by a diet of brilliant refereeing on the TV of the “near-perfect match”, they can cry foul, and, on occasion, forget what the R in TREDS is all about.
The progress in technology and its usage has been one of the great developments of the last twenty years, which also happens to be when rugby union turned professional. However, whether directly or indirectly, this has also coincided with a greater cultural emphasis on accuracy; everything should be demonstrably accurate and unbiased. Often in a data environment, artificial intelligence is the way forward, because it removes the need for humans, or so the narrative goes. In the same way that it’s possible to have the perfectly refereed match, it’s possible to have the perfect attribution model, the perfect active-learning segmentation and even an implementation that matches perfectly with other systems. But the true data professionals understand that there is no perfect game, and the key is in fact communication, not total accuracy. As one client put it recently when we discussed how two similar data systems would never correspond exactly, but the objective was to get within tolerance levels, “I’m glad you get it – now, if you could just help me explain that to the board, that would be great”.
The best data analysts are similar to the best referees. They understand that there is no perfect game, and the need to constantly explain what is going on with the data so their end users can play within the rules (or the constraints) of the available data systems, whilst answering the right business questions (both defensive and offensive ones) and hopefully enjoying the experience along the way. At the same time, they also need to ensure that their key supporting stakeholders on the touchline (the board, other departments and sometimes external suppliers) understand how you want the game to be played, so that they can also see what they need to see in the match, but can also adhere to your core set of values. They should thereby afford you the respect to get on with it, and not interfere on the data field, as it were, but to do so, you need to keep them informed.
In a world where more and more stakeholders will want to get more out of data, by all means use Machine Learning tools to help you with part of the game, but don’t lose sight of the fact that human-focused communication is the best way to ensure the wider data game is interpreted and understood in the right way.