James Taylor, CEO of Decision Management Solutions and a leading expert in helping companies adopt digital decisioning.
The concept of human-governed AI is becoming increasingly important as its use becomes more widespread. I believe that as AI and machine learning become more powerful and capable of making decisions, it is essential that humans remain in control.
Human governance of AI means that humans are ultimately responsible for the decisions made by these tools. Especially for business applications, human governance helps ensure that AI is making judgments that are compliant with regulations, ethical and respectful.
It is important to recognize that human governance cannot be applied simply by injecting a human into the decision-making process. This is sometimes known as the “human airbag” approach where concerns about AI are mitigated by putting a human in the loop to take the blame. In fact, to effectively incorporate human governance into AI, we must design the decision-making process explicitly to incorporate both decision-makers.
AI Risks And The Benefits Of Human Governance
Without human governance, AI may expose your company to unmanaged risks. In contrast, human governance of AI increases the likelihood that the AI will be trusted by customers and employees, making them more willing to use it.
Legal And Compliance Risks
The most obvious risk of not having human governance of AI is that the AI may make decisions that are not compliant or legal. The data that was used to train the AI might imply that a particular decision is the “best” but that decision may not be allowed. The AI may breach published policies, procedures, contracts or regulatory guidelines.
Decision-Making Risks
Some approaches to AI, such as generative AI or large language models, are already famous for making up plausible-sounding explanations that aren’t true. This is an obvious decision-making risk. Even if an AI does not make anything up, it may still make decisions that are not the best or most appropriate ones. This could be due to a lack of context or because the data that the AI is trained on lacks essential elements.
Acceptance Risks
Lastly, it is difficult to get people to accept AI if they feel like they have no control or ability to govern what the AI does. This may limit the potentially positive impact of AI in a business domain by limiting adoption or it may cause a customer backlash if a company starts to use it.
Best Practices In Human-Governed AI
Combining what AI and machine learning can infer from your data with the expertise of your staff allows you to impose guide rails, constraints, product designs, regulatory frameworks and policies. Ensuring human-governed AI means focusing on the decision as a whole—on a wide array of technologies beyond AI—and on continuous improvement.
Decision Thinking, Not Algorithm Thinking
To achieve human-governed AI, it is important to move away from thinking about AI in isolation. Instead, you should try to think about decision-making holistically—using machine learning and artificial intelligence within the context of a business decision. This means hiring people who can understand the decisions being made, capture the business rules that guide and constrain it and see exactly where within that context machine learning and AI is being applied.
Mixing And Matching Technologies
AI does not stand alone. Successful, human-governed AI is mixed with more “traditional” machine learning and analytics to create an automated yet governed decision. This means breaking down the decision-making into pieces and strategically assigning them to AI. It also means assigning the pieces that should be made based on human expertise to human decision-makers—not by bolting them on at the end, but by having their role be fully defined with specificity in instructions.
Continuous Improvement
Finally, as the term implies, the most important aspect is that humans are in the loop. Business experts watch the decision-making, assess its effectiveness and improve it using the data that is generated as the decision is made. Such a continuous improvement cycle allows you to refine decision-making in the face of new data and real-world outcomes. Expertise, judgment and results are injected back into an overall well-structured decision.
Towards Human-Governed AI
As we move into a new technological age, I see the growing need for human-governed AI. It can help ensure that humans remain in control of the decisions made by AI systems while still allowing us to maximally exploit AI in our business systems.
By understanding the decision-making process, capturing the business rules that guide and constrain it and showing exactly where in that context machine learning and AI are being used, companies can ensure that their AI systems are properly governed.
Embedding human-governed AI can help keep ethical considerations in mind, monitor and evaluate for potential risks and adapt and respond to changing contexts. In the end, I believe only human-governed AI will ensure that a company’s AI systems are properly governed and trusted by customers and employees.
Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?
Read the full article here
