Designing for the Senses: Epilogue

Intentional Delay

by Christian Cantrell

Part Three: Aural History

Several years ago, I created a digital version of the game Reversi (more widely recognized by its trademarked name, “Othello”). The most interesting thing I learned from the experience didn’t come from the design or development processes, but from watching users play against the AI. In order to conserve CPU cycles (and hence battery life on mobile devices) rather than calculating every possible move, and virtually playing each scenario out to its inevitable conclusion, the computer simply prioritized moves that were most likely to lead to capturing the most strategic positions on the board, then let the rest of the game take care of itself. The result was an automated opponent that was both fairly competent, and extremely fast.

The most interesting thing I learned about my Reversi experiment came from watching users play against the AI.

It was so fast, in fact, that it appeared to function instantaneously. Early in the game, when moves were limited and only small numbers of disks changed hands, players didn’t really seem to mind. But as the game matured and got more complex, with larger numbers of disks being captured with every move, I noticed players getting increasingly irritated with the AI — not because it was necessarily all that good, but simply because it was fast. The longer human players needed to decide on a move, the more annoyed they were when the computer instantly countered, robbing them of the satisfaction of ever really feeling like they were in control — even temporarily. Several of the matches I watched ended with the human player giving up in frustration well before the game was over, proclaiming that the computer was impossible to beat (it was not), and sometimes even characterizing the AI as arrogant. One test subject reported that, rather than enjoying the experience, playing against the AI made him feel stupid.

The “fix” I came up with was to introduce a delay that roughly corresponded to the number of disks that were to be captured. In other words, the more disks the computer was about to flip, the longer it would pause, as though the burden of coming up with such a dramatic move required noticeably more thought than lesser moves. I then refined the heuristic by adding another delay that corresponded to the number of disks that the human player captured on the previous move in order to simulate the computer being momentarily thrown off its game by its organic opponent’s unexpected guile. And finally, I introduced a condition that occasionally made the computer choose a move that was not necessarily the best possible option on the board so that the player might feel as though he or she noticed something that the computer — obviously somewhat flustered — had missed.

In other words, I introduced a touch of humanity into the AI.

An example of an AI without much humanity. (Spoiler alert: it didn’t work out well.)

As I expected, three things happened. The first two were things that a programmer almost never strives for: the computer got thousands of times slower, and the AI occasionally screwed up. But the third thing was something that almost every designer and engineer desperately wants: the experience of engaging with the program significantly improved. While players couldn’t get in as many matches during a session as they could before — and theoretically, they probably weren’t improving their skills to the same degree — they were enjoying their interactions with the computer far more, and in doing so, transitioning from test subjects to genuine fans.

My Reversi experiment has served as an allegory for me many times over the years as I’ve continued to study the factors and dynamics that make people feel affection, apathy, and even aversion toward their devices. For instance, after buying my first Mac mini, it helped me make sense of the completely impractical decision to put the SD card slot on the back where it’s almost impossible to reach instead of on the front (where, by the way, one or two USB ports would have been welcome additions, as well).

Wouldn’t it have been much more practical to put a few of these ports on the front where people could actually get to them? Yes. And no.

The answer is that only a relatively small percentage of Mac mini owners will ever use the SD card slot, so why ruin the coherent visual presentation of the front of the machine for everyone? And even those who do occasionally use the slot probably appreciate the clean and elegant design of the Mac mini far more than they resent occasionally having to reach around to the back of it and spend a few seconds groping. After all, most of us are perfectly happy to overlook the imperfections in both the people, and in the things, we truly love.

Christian Cantrell is a science fiction author, freelance tech writer, and Engineering Manager on Adobe’s Experience Design Team. Follow him on Twitter.