In response to my post on The Four Steps of Purposefulness, Robert Campbell sent me some interesting comments via e-mail. What follows is his comment and my reply.

On Tue, 1 Oct 2002, Robert L. Campbell wrote:

In Artificial Intelligence jargon, what makes step 4 hard is called a “blame assignment” problem. When you fail at some project, is it because your overall goal was unrealistic? Or was the overall goal OK, but one or more subgoals weren’t? Should you work harder–or work smarter–or change your overall goal–or abandon it? One thing that I think you can get AI and robotics types from every school of though to agree on: there is no algorithm for resolving most blame assignment problems.

Thanks for giving me a name to put with the error!

The basic problem of blame assignment, I think, stems from the fact that we are attempting to reach valid inductions under the worst of conditions. We often can’t afford to repeat our mistakes in a quest to discern the exact cause of failure. Additionally, we are often dealing with such complex situations (particularly when multiple other people are concerned) that we cannot effectively test whether X or Y or Z is to blame because so many of the variables are changing all the time without our even knowing.

Such problems do make the blame assignment problem inherent difficult. But for a great many people (particularly those that would benefit from Branden’s book), such problems of induction are never even encountered. These people are mired in their own beliefs about the rightness of their action, so they simply keep repeating the same bad strategy over and over again. For such people, to have the “blame assignment” problem would be a big step in the right direction!

Nevertheless, I do wonder what sort of methodology would be most effective in dealing with the inherent problems of blame assignment. There may be no single overall methodology, as situations may vary so greatly in their risks and complexity as to require radically different approaches.