Model-based reasoning is the use of a working model and accompanying real-world observations to draw conclusions. It plays an important role in artificial logic systems as well as reasoning in the sciences. The creation of the model is the time consuming aspect of this approach, as it is necessary to make the model as deep, complex, and detailed as possible to achieve the best results. Once a working model has been established, it may also require periodic updates.
In an example of model-based reasoning, a company could develop a working neurological model of the human body. The model would normally include information about the network of connections found in the central and peripheral nervous systems. Data about symptoms of neurological problems could be built into the system, using observations to create a matrix of known information. A user could potentially interact with the model by inputting patient symptoms, like slurred speech and unequally dilated pupils, and it would return a potential diagnosis, like stroke.
Such systems can have a wide range of applications in the sciences. Artificial systems can allow researchers to explore and test hypotheses. Model-based reasoning can also be the backbone of a monitoring system that sends alerts based on inputs. Climate modeling, for example, allows computers to take information about current weather conditions and run it through a model to provide information about budding tropical storms and other meteorological events of concern. Automation of some tasks can allow researchers to focus on other topics that require more complex reasoning.
The same concept can also underlie some forms of scientific thought. Researchers maintain working models about scientific concepts, like how tectonic plates work, and make observations to strengthen the model and develop a compendium of supporting information. This allows them to draw conclusion about scientific events, based on what they know from the model and the observations they have made. If, for example, researchers are monitoring a volcano, the model-based reasoning can allow them to issue an evacuation warning if the volcano's behavior is consistent with an imminent eruption.
Developing models can take time, patience, and input from a number of sources. The more points of data, the more accurate and detailed model-based reasoning can be. This can help modelers avoid potentially costly errors, like failing to anticipate an issue that would have been apparent with more data. As observations come in, they can be added to the body of knowledge, which may result in shifts to the model. For example, an observation could prove that a rule based on the model is actually incorrect, or does not account for a particular variable.