Once in hand, there are several mechanisms an organization can engage to ensure its principles are reflected in its day-to-day practices.
Policy: This Is Our Approach
Policies designate “a course of action.” Look at a definition of “policy” and you will find references to plans, strategies, approaches, and blueprints. Done correctly, policies chart the course between principles in theory and a plan of execution to put principles into practice. A well-crafted policy will:
- Clarify expectations: Clearly state how individuals are expected to engage and detail explicit outcomes or activities that they are responsible for or accountable to.
- Designate authority: Define who has the right to decide what. Whether it’s a governing council, an executive, or an individual contributor, policies should clearly delineate the boundaries, responsibilities, decision rights, and intersections between stakeholders.
- Provide guidance on priorities: As previously noted, principles do not always operate independently of each other. Policies govern what takes precedence and how a tie is broken when conflicts arise.
- Encourage progress over perfection: The best policies, in my experience, favor progress over perfection, thereby avoiding the perennial governance chicken-and-egg dilemma—namely, avoidance of codifying a policy due to lack of current compliance. Compliance, in turn, is hampered by a lack of policy to make the requisite actions so. It’s an easy trap to fall into.
Good policies go wrong when they, in essence, do none of the above. Your policies are likely more performative than productive if they:
- Prioritize perfection over progress or expect to “go green” overnight
- Do not clearly identify who is responsible for what
- Do not anticipate change and provide a mechanism for evolution over time
- Are not enforceable or enforced
Once you have defined what your organization stands for as well as why and when individuals are expected to engage, the next question is: How?
Standards: Consider This
A standard is an “idea or thing used as a guide, norm, or measure for comparison.” Look for definitions of the word “standards,” and you will find references to guidelines, benchmarks, measures, and criterion. As the definition suggests, standards are not one-size-fits-all. Depending on the problem at hand, the relevant standards may incorporate discrete methodologies, design metrics, guidelines, best practices, processes, and/or performance metrics. Organizations such as the professional association IEEE (among many others) provide the gold standard for standard definitions (pun intended) and have been actively engaged and leading the charge on creating comprehensive standards related to data and AI.
So, what are standards intended to accomplish?
- Increase literacy and understanding: They identify norms and generally accepted processes and procedures.
- Help you cover your bases: They provide practitioners with a level of confidence that all applicable dimensions of a problem have been considered and all requisite tools to complete the job have been accumulated.
- Establish definitions of quality: They provide objective and subjective goals by which progress, maturity, and completeness of execution can be assessed.
Processes and Procedures: Take These Steps
Processes dictate “a series of actions to be conducted in a certain order or manner.” Other synonyms may include procedures, techniques, routines, or methods. Some definitions of a procedure are interchangeable with that of a process. However, they are more commonly used in conjunction with one another (i.e., processes and procedures). In this case, processes dictate a series of steps or sequence of events to achieve a global outcome. Procedures document the steps required to achieve a specific activity/task described in the overarching process.
Processes and procedures:
- Instantiate best practices: Memorialize lessons learned and best practices so they are documented and available for everyone to learn from.
- Guide consistent execution: Provide blueprints to ensure that the like tasks are executed in a like manner utilizing best practices.
- Increase confidence in execution: Create a reasonable level of assurance that a task has been satisfactorily completed.
- Ensure quality outcomes: Ensure that tasks are executed to an agreed upon standard of quality or completeness. And that an outcome is of sufficiently high quality to address the need at hand.
Methods: Do It Exactly This Way
A method is a procedure focused on a very specific task. Synonyms include techniques, routines, formula, and mechanisms. In the realm of AI, a process might specify that teams evaluate fairness. Methods will exist for a spectrum of fairness metrics. For instance, equity (i.e., delivering equal outcomes) and equality (i.e., providing the same resources or opportunities) are two different concepts of fairness. And, therefore, by definition, they employ different methods for evaluation.
Regulations: Because I Said So
So where are regulations in this stew of governance mechanisms? Regulations are “rules or directives that are made, and typically enforced, by an authority figure.” Laws, statutes, and bylaws are commonly understood forms of regulation. As the definition implies, regulations are sticks, not carrots. As opposed to guiding or suggesting behavior or action, regulations dictate them. And, because the difference between carrots and sticks is the existence of penalties, violating a regulation typically comes at a price. These are most typically, financial penalties or the ability to operate as a business entity. Internal regulations carry similar penalties that include fines, loss of authority, demotion, or even loss of employment.
That said, regulations are not always discrete directions (e.g., if A, then do B, or don’t do C). While regulations can invoke explicit rules, they may also invoke the use of specific standards or methods for compliance. In other words, regulations can define process and procedural requirements as well as discrete actions or outcomes.
Regulations, should, in theory, be influenced by the espoused principles of the ecosystem (e.g., the industry, association, state, nation, or society) they operate within. However, it is common to find that regulations lag societal and public expectations of what is right and wrong. And, while current regulations, including the proposed EU European Commission’s Regulatory Framework, aim to proactively avoid harm, legality often follows only when harm has already been experienced.
What’s Ahead: Making Good Governance Work Is No Joke
Poor governance can be quite farcical. Good governance must engage and guide a diverse set of stakeholders with a correspondingly wide range of responsibilities. As a result, there is no single governance mechanism—be it a person, principle, policy, or procedure—that carries the day when it comes to comprehensive, ambient enablement.
Governance must be present but also contextually relevant and tuned for execution at all levels of the organization. This entails an intertwined web of enabling mechanisms from high-minded principles down to nitty-gritty methods. Miss any piece and your initiative will likely flounder. Somewhat similar to my feeble, and ultimately unsuccessful, attempt to craft a joke worthy of this article’s title.