September 25, 2008

Principles are Coming

Principles are Coming

More judgment and knowledge needed

While the finance industry is moving toward more rules and exceptions, and rules-based regulation, financial accounting and reporting is moving in the opposite direction, towards fewer rules and exceptions. Accounting and tax is moving towards more principles-based regulation.

The US is moving away from GAAP (Generally Accepted Accounting Principles) and towards IFRS (International Financial Reporting Standards). IFRS is the reporting framework used by most of the world today, and it has growing support in the US. IFRS relies on professional judgment rather than detailed rules. Under this principle-based approach, management will have a mandate and obligation to exercise its own best judgment when making decisions.

Companies will need to implement systems that use knowledge and judgment to make principle-based decisions.

It is time to adopt knowledgebase technology and knowledge management. It's time to build knowledge bases and embed knowledgebased technology into operations and existing systems.

Knowledgebased systems that are engineered and architected properly can

  • follow principles and guidelines
  • automate management's best judgment
  • ensure compliance
  • and deliver trust. These expert systems can be trusted because they use expert judgment to make the same decisions top experts would make, thus improving the quality, accuracy, and consistency of decision-making.

I don't believe there is any other practical or proven way of automating human judgment, other than building intelligent, knowledge-based systems.

Knowledgebased systems are the solution for principle-based compliance.

April 05, 2007

Why business rules? Why not expert systems?

James Taylor over at Fair Isaac has a really good list of "Why business rules?" I agree with most of the points, except the stuff about expert systems.

Maybe the question should be "why not expert systems?"

The dirty little secret is that a lot of the rule engines out there were originally called "expert systems" or "inference engines", then they were called "business rule engines", and today they are known as "business rule management systems. (See the business rules hype cycle)

Of course, everything is better today. And faster. And connected. When expert systems first came out, the Web wasn't even born yet, and PCs were running at 10 mHz. 

The biggest problem we had at Mobil Oil between 1988-1994 when we were building the Global Expert System Strategy and Lube Knowledgebase Strategy was making and mailing floppy disks to all our affiliates.

I remember one day we were showing the customer (an executive in Mobil Marine division) a demo of the expert system, his comments were:

  1. This is like an intelligent checklist, it never asks un-needed or dumb questions!
  2. I like that I can click on an underlined word (a hyperlink) and popup a definition, photo, go to the next page, or whatever!
  3. This is not like our other DOS or mainframe apps. Our users will not like the fact that this works on a "one page at a time" metaphor,
  4. because we're forcing users to fill out information or answer questions on the page (screen), then they have to press enter to go to the next page (screen).

That one page at a time metaphor he described was basically how the World Wide Web works. We were doing this in a business rule engine (BRE), aka an expert system (ES) in 1988. Before WWW. Before Windows.

(Want proof, go here and click on the photo on the right. There's a picture from back then, in my younger days... the program on the PC behind me is 1DirPlus or something like that.... Definately B.W. Before Windows). And so back then we built expert systems that did reasoning, chaining, hypertext / linking, and of course inferencing. Basically they would fire rules exactly the same way a modern rule engine would today. And give the same answer the expert would give,

Even after the experts retired long ago!

We did that in AION. We could have used Neuron Data (which evolved into Fair Isaac Blaze Advisor), or we could have done it in ILOG. Or any number of other ES tools at the time. Some of them are still around today. (See BRE Family Tree)

Distribution of expert systems, and access, is one of the reasons they "never took off". People used to say expert systems were a solution looking for a problem. Deploying expert systems on the web solves those problems.

I think the Web is "the problem" that expert systems were looking for. The Semantic Web is reigniting a lot of the good stuff from the AI/ES days. Adding intelligence and reasoning to applications is what expert systems have been doing all along.

And by the way, not everyone agrees that expert systems never took off. I certainly don't.

As Richard Barfus, CEO of MindBox, (an ES/BRE/BRMS firm) likes to say, "Expert systems didn't really go away. They went undercover."


October 24, 2006

What's The Greatest Software Ever Written?

Now here's an interesting story by Charles Babcock in InformationWeek, Aug. 14, 2006:

"First, let's set criteria for what makes software great. Superior programming can be judged only within its historical context. It must represent a breakthrough, technical brilliance, something difficult that hadn't been done before. And it must be adopted in the real world. ...

The AI application that produced the first real breakthrough was the inference engine, a system with a knowledge base of conditions and rules. Such a computer can match a condition, such as a 104-degree fever in a patient, to a rule, such as the fact that bacterial infections cause high fevers. One of the best, the Mycin medical diagnosis system, could correctly identify bacterial infections in people based on their symptoms 65% of the time. That's better than most nonspecialized physicians. But it never moved out of the lab into popular use. No one knew who to sue when it was wrong.

My favorite AI package was IBM's Deep Blue program, which defeated chess Grand Champion Garry Kasparov in a six-game match. ... AI software can be impressive, but all my examples fall short of being among the greatest. ... Continuing into modern times, Google, in one aspect at least, represents great software. ... American Airlines' Sabre system was great, showing how software could evolve beyond the tactical needs of business and into the strategic. Sabre had the ability to match a customer's travel needs with the flights available at a travel agent's office. ... So how do I rank my candidates on a list from 1-12? In descending order, the greatest software ever written is: ...."

For the full story, click the link below:

Source: August 14, 2006: What's The Greatest Software Ever Written?

September 27, 2005

M&A: Fair Isaac acquires RulesPower

Sep. 27, 2005: The Business Rule Management industry shakeout continues.

Today Fair Isaac announced it has acquired RulesPower, development staff, rights to the RETE III algorithm, and other patent pending technology.

More details on strategy & direction to be announced later by FI. See for details.
Locations of visitors to this page