Recently a few colleagues and I had a short discussion on models covering human performance and on how to improve them. It was an interesting conversation and I decided to make some core ideas available here.
We depend on models and we need them in teaching and in consulting.
The tricky question is what does it mean to improve models: What does BETTER mean?
This question can only be answered out of a specific perspective and therefore it is context specific.
Let’s take a look at the known (H)PT models.
Performance improvement claims to take a systemic approach. This entails that models have to show the relevant elements and their relationships (How do they interact with each other?). Let’s ignore for a moment that relevance is context-specific also. Let’s simply look at the known models and check if they show elements and their relationships.
Except for Geary Rummler’s and Dale Breathower’s very simple Job Performer Model and the models we developed (in discussion with Geary) I don’t see many out there that show elements and relationships. (A process flow like analysis, design, implementation, evaluation does not show relationships between the elements that influence performance).
What people usually call a model is a list of elements that they somehow cluster. Good examples for that are Gilbert’s work and the Six Boxes built on Gilbert.
These models do not meet the standard of being systemic. Improving them would simply mean showing how these elements relate to each other. Show arrows connecting the elements and displaying type of influence. Would be a huge improvement already.
When it comes to the details it gets tricky again. How do these elements influence each other, in what sequence and what is the strength of this influence? The answers will be -guess what – context-specific.
Don’t get me wrong. I’m not against models. I’d only argue teach people how to develop models, make them understand what are the strengths and limitations and how to use them.
Other than that we have long lists of elements that influence performance. They can be helpful starting points in developing context-specific models. Pick from these lists and start developing a model that is useful for the problem you want to solve.
Making generic models is also not wrong because they can serve as starting points also.
Only the idea of BETTER is one that might be futile.
Improving existing generic modeIs will end up in adding ever more elements and connections which means adding complexity. Complexity adds a ton of additional problems. You might end up with something that looks good on paper but only will become meaningful if you reduce it to what is relevant in a given context.
Then what is BETTER? Building a model by identifying relevant elements and their connections based on a list or building a model by distracting elements and connections from an overly complex generic model? Or a mixture of both? Building better models is tricky.
And we have not discussed the issue yet that models are not theories. Models are a simplified incorporation of a theory. In any case, the more interesting progress comes from improving the theories behind the models.