Wednesday, February 10, 2010

The Toyota Problem, Interaction and Complexity

I guess everybody has heard about the Toyota car problem. This is a terrible situation for Toyota, of course. At the same it also exposes some interesting interaction issues. Without knowing anything about what has really happened and how much Toyota really knows, I think this is a kind of situation we might see more of in the future.

When artifact complexity increases as a consequence of growing features "need", new problems emerge. The blend of mechanical/physical material with digital material seems to be part of the Toyota problem. A spokesperson said something about "not being able to locate the problem". A modern car is highly complex and is an example of an system where parts "collaborate" in a highly intricate way. So, is the gas pedal issue a mechanical problem or a digital problem. According to an (TV) expert, Toyota is hoping and praying that it is a mechanical issue, since it will be possible to locate and hopefully add or remove some mechanical parts. If it is a digital issue then "locating" becomes increasingly difficult. Maybe the problem only happens with a highly improbable occurrence of combination of system states, mechanical as well as digital. It might actually be an untraceable problem. And it might be impossible to test. No usability testing, no matter how comprehensive, might catch or isolate the problem. However, every day millions of drivers round the world conduct a full scale usability experiment, in rare situations with drastic consequences.

As a coincidence, my colleague Lars-Erik Janlert and I, will have an article in the ACM ToCHI in a few months where we explore interaction complexity and actually use the brake system in a car as a core example. I can't see any simple solutions or obvious things to learn from the Toyota case. Well, there might be one systems that have reduced complexity so they are possible to comprehensibly test.

In our article we discuss how less system complexity usually lead to increased interactivity complexity. For instance, ABS brakes move control from the driver to the system, since the brakes almost becomes an on/off digital button while the mechanical brakes relies on the exact (analog) pressure from the drivers foot. So, there is a design choice---safety in form of brakes that can adjust to the surface or safety in form of a brake system that is less complex? That is where we are today with many of our artifacts.

As a side note, I wonder how often old cars had mechanical problems that led to serious accidents? Maybe the problem is not that today's cars are less safe, of course they are not, but that the acceptance of malfunction is being reduced to a ridiculous low level.

1 comment:

Olle Hilborn said...

A comment on that side note: Yesterday I was driving home my car, a Volvo 855 -95, so an old car with mechanical brakes and mechanical gas control. I was driving on to the highway from a standstill so I pushed the gas pedal down to to the bottom and when I released it, it was still stuck. I solved the problem by jerking the pedal with my foot and it came loose after just a couple of seconds, so it was no big deal really.

I agree with you that it might be that we are less tolerant to malfunctions today. But in my case it was no problem because of the design of the whole car; it's really not a fast car and accelerates pretty slow. If I were to drive one of today's cars who carries a lot more power under the hood it might have been a problem. Just an example on the classic problem with changing one part of a design can alter the whole way it is used =)