Immanuel Kant, Data Modeler

Only the users can say whether a data model accurately reflects their business environment. What happens when the users disagree among themselves?

What if one user says orders have a single salesperson but another says that sales teams produce some orders? Who is correct? It's tempting to say, "The correct model is the one that better represents the real world." The problem with this statement is that data models do not model "the real world."

A data model is simply a model of what the data modeler perceives. This very important point can be difficult to understand; but if you do understand it, you will save many hours in data model validation meetings and be a much better data modeling team member. The German philosopher Immanuel Kant reasoned that what we perceive as reality is based on our perceptive apparatus. That which we perceive he called phenomena. Our perceptions, such as of light and sound, are processed by our brains and made meaningful.

But we do not and cannot know whether the images we create from the perceptions have anything to do with what might or might not really be. Kant used the term noumenal world to refer to the essence of "things in themselves"-to whatever it is out there that gives rise to our perceptions and images. He used the term phenomenal world to refer to what we humans perceive and construct. It is easy to confuse the noumenal world with the phenomenal world, because we share the phenomenal world with other humans. All of us have the same mental apparatus, and we all make the same constructions.

If you ask your roommate to hand you the toothpaste, she hands you the toothpaste, not a hairbrush. But the fact that we share this mutual view does not mean that the mutual view describes in any way what is truly out there. Dogs construct a world based on smells, and orca whales construct a world based on sounds. What the "real world" is to a dog, a whale, and a human are completely different. All of this means that we cannot ever justify a data model as a "better representation of the real world."

Nothing that humans can do represents the real, noumenal world. A data model, therefore, is a model of a human's model of what appears to be "out there." For example, a model of a salesperson is a model of the model that humans make of salespeople. To return to the question that we started with, what do we do when people disagree about what should be in a data model? First, realize that anyone attempting to justify her data model as a better representation of the real world is saying, quite arrogantly, "The way I think of the world is the way that counts."

Second, in times of disagreement we must ask the question, "How well does the data model fit the mental models of the people who are going to use the system?" The person who is constructing the data model may think the model under construction is a weird way of viewing the world, but that is not the point. The only valid point is whether it reflects how the users view their world. Will it enable the users to do their jobs?

Discussion Questions

1. What does a data model represent?

2. Explain why it is easy for humans to confuse the phenomenal world with the noumenal world.

3. If someone were to say to you, "My model is a better model of the real world," how would you respond?

4. In your own words, how should you proceed when two people disagree on what is to be included in a data model?