IBM using math models of employee behavior to automate management

businessweek.com

14 points·ryanwaggoner··4 comments

This model can only be as good as its data. Which means more forms to fill out. As any IBM cog can tell you the level of that is already staggering. The only way I can imagine this big brother stuff helping is if they use it to root out the bad apple consultants that eventually find their way to every successful project weighing it down until it implodes.

From the sound of it, they're mining data that exists as a byproduct of existing processes, not asking for new forms.

Augment management is probably closer to the truth than automate.

Judging from a lot of the management I've observed, even augment is probably a bit optimistic. I can see several issues with this approach:

1. The article paints a picture of a world where "management" basically becomes a role of data-entry and adjusting some parameters and constraints until the system tells you what to do. It seems unlikely that managers are going to support such a system.

2. The system can only do so much to take the human factor into account. The author of the article compares this system to efforts by IBM over the last few decades to apply similar research efforts to supply chain management. But people aren't widgets in a global supply chain operation, and are unlikely to appreciate being reduced to an interchangeable commodity deployed by an algorithm.

3. The system is not allowed to use personnel files, including performance appraisals, in the analysis. It would seem like this would be one of the richest sources of actual data in the system.

I still think the approach has some value and is definitely interesting, but I'm not sure that it will have the impact that they seem to expect.

Reply on news.ycombinator.com