Dealing with Imperfection
Imperfect rules will result from imperfect information. Davide Sottara, a research scientist at the University of Bologna, illustrated the challenges of reasoning with fuzzy information. Side note, University of Bologna has the distinction of being the oldest continuously operating university in the world and a wonderful place to visit. Sottara referenced a well known story by James C. Bezdek, distinguished professor and researcher known for his work on fuzzy systems. In a nutshell, sometimes we face the following challenges:
- lack of information
- ill-defined information
- erroneous information
Sottara emphasized that fuzzy information deals with “imperfection” rather than “uncertainty.” Imperfection is, in essence, the opposite of perfect precision. As a result, imperfection brings us outside the scope of what traditional rules engines do. However, removing imperfection from the picture (by ignoring it or simplifying it) leads us to ignore essential information that’s relevant for writing rules.
More Robust Systems
Sottara’s work focuses on creating more robust systems that may be applicable to imperfect information yet still leverage rule based systems. Using a modus ponens model, he illustrates the impact of introducing imperfection to the way the rules engines work. In essence, we need to think about the connectors (combination of premises, logical implication, etc…) differently, solve conflicts created by the imperfection, and handle missing values. Davide would like to preserve the structure of rules and hide those complexities from the rules expression itself.
Sottara compared and contrasted various approaches to deal with imperfection: frequentist, Bayesian, and Fuzzy logic. Each one has a different philosophical foundation and application domain. The main takeaway from Sottara’s talk is that you should not forget about imperfection when creating and managing rules.
Food for thought. Charles Young asked what I think is a very good question: Why did rules solutions in the early days include Bayesian or equivalent analytics capabilities but not anymore?
See our recap on Jacob Feldman’s talk
Learn more about Decision management and Sparkling Logic’s SMARTS™ Data-Powered Decision Manager