Is the shuffle feature really random?
I have an extensive music collection which I carry around with me on a handy 3 inch by 2 inch box of tricks in MP3 format. Even twenty years ago it would have seemed inconceivable that I could carry around my complete record collection and access any track at the click of a switch but I can and, perhaps, life is the better for it. What I find I use increasingly more often, because choice requires thought and concentration which in turn distracts me from doing what I otherwise would be doing, is the shuffle feature on the player. Start off with one track, hit the shuffle button and let the machine’s algorithm select which tracks to play for your listening pleasure. What could be easier?
What I find though is that the same tracks appear from time to time whilst other tracks may as well not be in the machine as they never see the light of day. This set me wondering how truly random the shuffle feature really is – a 21st century concern, if there ever was one.
Of course, behind this rather diurnal concern is a bigger issue – what do we understand by randomness. If we are expecting every track on our MP3 player to be played just once but in an unanticipated order then this is not random. For there to be true randomness at play then it is likely that the same track or tracks will pop up within the sequence, albeit distributed randomly. The problem is that what is really random may not seem to be random to us.
What we are exposed to, according to those that know about these things, is a phenomenon known as Gambler’s or the Monte Carlo fallacy. Simply put, this is the belief (mistaken as it is) that if something happens more often than normal during a period, then it is less likely to happen in the future and if something hasn’t happened in a particular period it is more likely to happen in the future. If true randomness is at play, then this cannot be the case because the event is just as likely to happen in the future or not as it did or did not in the past. The consequence of this fallacy at play is if we hear a couple of tracks by the same artist in short order we think there is a problem with the random generator.
Although this makes the phenomenon of the recurring shuffle items more comprehensible it is not quite as simple as that (surprise, surprise!) because we are at the mercy of inherent deficiencies of the algorithm deployed to generate random numbers. If you think about it, computers are programmed to be consistent and to deliver the same (correct) answer. In other words they are deterministic, a characteristic which is diametrically opposed to randomness.
Often, in order to give the sense of randomness, programmers will use an algorithm using a pseudo random number generator which starts with a common seed number and follows a pattern. The results are often sufficiently complex to make the pattern difficult to discern but because the algorithm is carefully defined and consistently repeated, they cannot be truly random.
And then, of course, we compound the problem by turning our MP3 player off because we have lives to lead and we can’t, pleasurable as it may be, spend all our time listening to our music. And this compounds the problem because many a shuffle feature will start again which means, depending upon how large or small your collection of tracks is from which it selects, you run the risk of hearing some of the tracks previously selected again.
So now we know!