In the new age of generative AI and its deeply statistical approach, it’s time for a major rethink of rules, and the many things they do for us. After all, we remain human, and we will always live and work in communities and groups of people. The first thing you find is that wrapped up in rules is important knowledge that can be passed on to others’ advantage, not least of which is to avoid risks. This discussion enumerates all the reasons for rules, not the least of which are fixing data quality and developing better requirements. Are rules front and center in your approach? They should be!
Extracted from Rules: Shaping Behavior and Knowledge, by Ronald G. Ross, 2023, 274 pp, Chapter 1.
In 2005, I broke two basic rules of responsible ladder usage:
- Never stand on the top rung of a ladder.
- Never use a ladder alone.
I paid a hefty price – a broken right arm and a stay in the hospital the night before a hurricane was supposed to hit my home town. (Yes, my ill-advised ladder usage and the looming hurricane were related.)
The incident was also a teaching moment. Ladders are more dangerous than you might imagine. You can bet I have never violated those rules since! (And now I can hope you won’t either.) Wrapped up in those rules is important knowledge that can be passed on to others’ advantage. Better to learn dangerous things through rules than personal experience(!).
Rules and Risks
As my ladder episode indicates, one fundamental purpose of rules is to protect against significant risks. They provide markers of known dangers, guardrails against counterproductive behavior. Good ones tend to keep you on safe, productive paths.
In a day and age of increasingly dynamic or unscripted behaviors, guardrails and rules of engagement are ever more essential. That’s all the more true for artificial intelligence (AI) since it has no innate biological or cultural constraints.
Rules and Community
See if you recognize the following children’s game from just the rules:
- Players must walk in single file around a group of chairs while music plays.
- Each player must sit in a chair when the music stops.
- A player not sitting in a chair when the music stops must be eliminated from the game.
- One chair must be removed each time the music restarts.
- A player is considered the winner of the game if they are the only player sitting in a chair when the music stops.
The game, of course, is musical chairs.
Can you imagine any game without a set of rules? Where there is a group or community of people, and it seeks some common purpose or goal, there will always be rules in some form. Rules are how you create order from disorder. That’s what laws, statutes, regulations, contracts, agreements, etc. are all about. Rules are also how you remember things and avoid unconscious biases and discrimination (perhaps in machine learning as well).
Rules and Governance
Policies are basic to the working of governance. Policies shape the broad contours of desired behavior within or across communities of people. Associated with policy are sanctions, either positive ones to reward desired behavior or negative ones to discourage undesired behavior.
Policies generally must be interpreted to apply at the fine-grained level of everyday group behavior. What do you get when you interpret policies? Rules. If you create policies someone – maybe you, maybe someone else – will also be creating rules. Along with penalties and incentives, policies and rules represent the gear box of governance.
Besides policies, institutional knowledge also involves norms (often tacit) and practices and procedures (sometimes formalized, sometimes not). These too can be rich sources of rules highly relevant to some or all of the community.
Rules and Innovation
These days innovation is top of everyone’s mind. Many shout mindlessly “break all the rules!” as if doing so will cure all ills of stagnation. Catchy slogan but like all panaceas, simplistic and misdirected. Here’s a better view.
Clear, predictable rules protect freedom of action as much as constrain it. Rules determine not just what is prohibited, but also help protect what is allowed. Rules that are explicit can be incrementally improved upon.
By contrast, failure to spell out what is and isn’t allowed can have a chilling effect, in which people steer clear of innovation. Few want to risk doing things differently knowing others have been sanctioned by unpredictable authorities for overstepping invisible lines.
In a great many cases, when people say they have too many rules what they really mean is that rules are implemented rigidly in processes and procedures that are overly complex. Or the rules are so poorly encoded as to seem voluminous. It all builds up like scar tissue. Yes, in that sense they are absolutely right – ‘too many rules’!
Rules and Strategy
Thinking about rules in the context of games is instructive in another sense. To win games, you employ strategy. What does strategy do? It identifies the ends (goals and objectives) you want to achieve and the means (including policies and rules) to achieve them.
But at some level, goals or objectives always conflict. If you go too far in one direction to satisfy some goal(s), you can easily undercut your ability to achieve some other goal(s). At its heart, strategy is about establishing optimal compromises between conflicting goals so as to ensure highest chances of success.
How do you express those compromises? By establishing policies and developing rules. Here’s a simple example. Suppose order fulfillment in a business has three business goals:
Goal 1: To stay competitive.
Goal 2: To keep customers satisfied.
Goal 3: To make a profit.
The business decides to accept orders on credit as part of its strategy. Doing so will help it (1) stay competitive and (2) keep customers satisfied. But there’s a risk: Non-payment of purchases made on credit. Since the business also needs (3) to make a profit, how does it protect itself? How far is it willing to go to achieve the first two goals at risk of the third? To resolve the conflict, it establishes the following rule:
Orders on credit over $1000 should not be accepted without a credit check.
The motivation for most policies and rules is seldom worked out as explicitly as this, but dig deep enough and you will always find risk and conflict resolution. (If you can’t, you might question why you need the policy in the first place.)
Rules and Customer Experience
The only thing worse than a consistently bad customer experience, is an inconsistently bad one. With a consistent experience, you expect it. When it’s inconsistent you don’t know what to expect!
How do you ensure consistent customer experience? Rules. Not just rules applied consistently at one touch point, but rules applied consistently across all collaborations and all channels (both digital and in-person) through which parties interact. And to the data on which all the interactions depend.
Rules and Concepts
As earlier, rules often pertain directly to behavior. They can also relate indirectly to behavior, by helping to shape the understanding on which behavior is based. Specifically, many rules provide definitional criteria. To illustrate, let’s return to the game of musical chairs. The fifth rule from earlier determines the winner of the game:
A player is considered the winner of the game if they are the only player sitting in a chair when the music stops.
The term ‘winner’ should have a definition, of course, but the rule above is almost certainly not part of that definition. The rule is supplemental to the definition, providing criteria to determine which person can be classified as the winner.
That’s a child’s game, of course, so the criterion is simple. But consider the complex concepts we deal with in real life – e.g., ‘market share’, ‘restricted air space’, ‘freedom of speech’, ‘money-back guarantee’, etc. Each of these concepts could have many dozens of qualifying definitional criteria.
Your knowledge is not complete until you make those definitional criteria explicit as rules. Otherwise, you will sometimes use the concepts incorrectly – not only disrupting business activity but likely producing bad data as well.
Rules and Inference
A special case of definitional criteria are rules that support what is commonly called inference, rules that infer new facts from existing facts. They too are an important form of knowledge-building rules. Here is a simple classic example:
If it is raining, then streets are wet.
The fact of wet streets is the logical implication of the fact of rain happening.
Aside: Just because an inference can be made in one direction does not necessarily mean it can be made in the other direction. For example, you can’t assume ‘If streets are wet, then it is raining’. A watermain might have burst, the city might be washing the streets, etc.
Such inference rules are the prominent feature of traditional expert systems (non-statistical AI). They are often strung together (chained) in order to make more sophisticated determinations.
In the past decade, machine learning has proven far more powerful for many forms of intelligent behavior than expert systems, so inferencing has fallen into disfavor for many kinds of bots. One important exception has arisen in business, however – the automation of operational business decisions using decision models and decision tables.
Actually, there’s no surprise there. You’ll always resort to explicit rules whenever you want both the following:
- Consistency of results no matter what the process.
- Traceability – being able to explain the ‘why’ of results.
Rules and Requirements
Rules provide the basic motivation for many requirements. Suppose I say, “We need a piano.” Would you have any idea why we need the piano? How about if I also gave you the five rules from earlier for the game of musical chairs. Then would you know why we need a piano?
The five rules, however, actually say nothing whatsoever about pianos. The rules simply call for music. Instead of a piano, you could simply play music from an electronic device. ‘Music-playing electronic device’ is probably a better, more practical requirement than ‘piano’.
See the difference between rules and requirements? Rules shape activity for groups and communities of people; requirements identify how you’ve decided to support that activity. The rules will remain relevant throughout the life of the activity; the original requirements much less so after implementation.
To use rules in automation projects, do you have to know all the rules upfront? Well, yes you should – if we’re talking about contractual or other legal constraints. Otherwise, of course not! You live and learn.
The goal is to develop the rules for an activity such that they are shared, retained, and accessible for the whole community. And, of course, revisable – because the rules will change over time.
Rules and Data
The connection between rules and data is actually simple. The results of business activity are recorded as data. So, rules for business are also actually rules for data. Not a different set of rules, but the very same rules!
If you want high data quality, ultimately the best and perhaps only way to get there is through rules applied in real time with respect to business activity.
Author: Ronald G. Ross, Co-Founder and Principal of Business Rule Solutions, LLC
Ron is one of the world's foremost authorities on policy interpretation, rules, concept models, business vocabulary, and data design. He is Co-Founder and Principal of Business Rule Solutions, LLC (BRS). At BRS, he has consulted to many hundreds of companies and government bodies around the globe.
Ron is the author of 12 professional books, including his most recent:
Rules: Shaping Behavior and Knowledge, by Ronald G. Ross (2023), on policy interpretation, rules in natural language, and governance.
Business Knowledge Blueprints: Enabling Your Data to Speak the Language of the Business, 2nd ed. (2020), on structured business vocabulary, definitions, and concept models.
Ron is Chair of Building Business Capabilities (BBC), the official conference of the International Institute of Business Analysis (IIBA®). Ron has keynoted dozens of conferences and given seminars to many thousands of people worldwide. He is currently Executive Editor of BRCommunity.com and its flagship on-line publication, Business Rules Journal.
Ron holds an M.S. in information science from the Illinois Institute of Technology and a B.A. from Rice University.