As digital systems have become increasingly pervasive in our daily lives, organisations feel an increasing pressure to adopt AI as quickly and as extensively as possible. Driven by the promises of increased productivity, stakeholders are expecting returns to their investments. This has led to sentiments that there is simply no time or resources to dedicate to governance.

But is governance truly halting innovation and making us lose opportunities offered by AI? Or is there perhaps more to governance that might turn it into an enabler rather than a barrier?

In this blog, I combine insights from research and my experience in AI industry to shed light on the role of governance and ethics in AI innovation.

We’re in a hurry and governance is slow. Or is it?

The current AI hype loud. Some parade the possibilities of general-purpose technology, and others paints dystopic scenarios where machines take over and humans perish.

When we take a close look at these claims (as several scholars did in this special issue in AI & Ethics,), the motivations are strikingly similar: they both paint a picture of a powerful technology that requires significant investments to fulfil its promises. As we argue in this article, these promises are often overexaggerated.

We need healthy optimism about technological development. However, cruel optimism can lead to misinformed decisions about AI, what it’s good for, and how it should be governed.

For AI governance, the hype brings significant pressure for public and private sector to use AI. It makes us think AI is an inevitability, and those who don’t act now will be left behind.

What we tend to forget is that AI governance is not about pulling the breaks. It’s about advancing to a desirable direction.

Systems that lack governance often do not fulfil the promises of valuable, life-improving, productivity-enhancing technology. Instead, we end up with lost potential, increased risks and harm.

Ethical governance enables innovation

Luckily, we don’t need to choose between speed and governance, because ethics and governance are shown to improve AI innovation. As Bednar and Spiekermann (2023) show, including ethics in IT innovation can surface more and richer value ideas. Industry reports give similar signals, as e.g. IBM’s Institute for Business Value surveyed 915 global executives found out that organizations that invest more in AI ethics achieve higher operating profit, stronger return on investment (ROI) and measurable competitive advantage from AI.

It is often the lack of governance that hinders innovation, as uncertainty makes it difficult to innovate with confidence.

The feeling that governance and ethics are blockers is a legitimate concern arising from lived experiences. Cumbersome governance processes and checks, unnecessary processes and hierarchy are issues that many have become to associate with AI governance.

This happens because governance is often an afterthought, not a strategic choice. It might be driven by compliance with laws and standards, which blurs the overall goal and vision of well-governed AI development and deployment.

AI governance models should be designed so that they

1. advance organisations’ goals,
2. support compliance with norms and values in a meaningful way, and
3. enable humans to flourish

That’s when AI governance becomes an enabler rather than a hindrance.

The missing piece is often ethics

Many existing governance models describe ethics as a set of principles that need to be operationalised. However, this means ethics is usually seen only as a constraint.

In truth, ethics is just as much about finding out what’s really valuable, what improves our lives and how to best reach positive outcomes.

Who wouldn’t want that out of AI, too?

Many scholars show that ethics is about asking very concrete questions during real-life AI development and adoption. For example:

ECCOLA-cards developed by researchers (including some from GPT-Lab) bring forth questions especially for developers:

• “Did you assess the broader societal impact of the AI system’s use beyond the individual (end-)users?”
• “Are the people involved with the development of the system also involved with it during its operational life? If not, they may not feel as accountable.”

Data Ethics Decision Aid (DEDA) developed by University of Utrecht researchers guides ethical considerations for multi-stakeholder settings:

• ”Could reusing outcomes as new input unintentionally cause a negative feedback loop, potentially reinforcing inequalities?”
• ”How is (data) sovereignty/independence pursued?”

REAL cards developed at University of Helsinki introduce broader considerations for algorithmic systems:

• “Who gets to decide what is replaceable?”
• “Does the system affect how professionals uphold their standards?”

Finding time and space for ethics can be challenging. When I studied ethics questions to ask during AI development, I concluded that we need structures to guide us in practical settings.

Effective AI governance establishes processes for both deliberation and action. Intentionally deciding what to build is just as important (if not more) than deciding how to build it.

In GPT-Lab, we want to ensure that organisations across sectors can make informed choices about what kind of governance works for them. We want to ensure that more organisations can experience AI governance as an enabler that helps humans and societies flourish.

We are actively looking for collaborators to work on even more. Don’t hesitate to contact us!