Need an "AI Ethics Committee?"

Nine steps to setting up an AI ethics committee in your organisation

I have a confession. "Committee" (and particularly its junior, interfering cousin "Sub-Committee") is a "trigger word" for me.

It conjures up images of bureaucracy, procrastination, and one particular memory of a late meeting that was followed by an overly formal "I propose that we" - followed by an "I second the motion" and an "all those in favour?"

It took all my strength of mind not to stand up and shout "FFS...there's only five of us in this room, let's speak like humans..."

And yet...if you’re using AI in business you might just need an AI ethics committee of some sort. Either because:

• There's no one person who is willing to take on the entire risk of the business suffering a failure as a result of AI (in house lawyers take note - this might be the alternative!); or
• Because AI projects inevitably have a number of stakeholders (legal, HR, business operations) who each have a different understanding of the risks and rewards of a particular project.

So -assuming you do need a committee - how do you get one together? Here are nine suggested steps:

Who do you need…??

There are a few key people you're likely to need on the committee:

• Someone from the Senior Executive: To endorse the committee and provide necessary resources (and make sure the committee's decisions are taken seriously).
• A Committee Chair: To lead the committee, set agendas, and ensure follow-up on decisions.
• Committee Members: Representatives from legal, IT, HR, and operational departments who can contribute their perspectives and expertise.
• A Compliance Officer: Someone who monitors adherence to ethical policies and reports to the committee (this may or may not be the in-house counsel).

What should they do?

1. Define the Committee’s Objectives and Scope

So first things first - the committee will need to determine the key areas of focus such as data privacy, algorithmic fairness, and transparency and to work out the roles and responsibilities of the committee. Ideally individuals would have defined roles, based on their expertise.

2. Select Committee Members

You're going to want to choose members based on expertise, influence, and perspective diversity. Aim for cross-departmental representation to include various stakeholder views.

3. Develop an Operational Framework

How often will the group meet? How will decisions be documented and communicated to the rest of the business?

4. Create an Initial Agenda

You can probably identify the immediate areas where AI ethics guidelines are needed - what are the important decisions that AI will be involved in? Recruitment? Sales? Marketing? Who will create the relevant policies for the most urgent AI apps?

5. Conduct Training Sessions

You can't just wing it. If you've not had to deal with ethical issues before its likely that the committee (and other stakeholders) will need some sort of workshops to get up to speed on AI ethics and relevant regulations. If you know any you might want to bring in external trainers* as needed to help with this.

6. Monitor and Evaluate AI Initiatives

What AI projects are currently in place? What sort of risk assessment has been carried out? You may find that you need to review these to align with established ethical guidelines.

7. Report Findings and Recommendations

One of the key deliverables is likely to be some sort of report for executive leadership - what's working? What isn't? One of the roles of the committee may well be to suggest improvements and provide feedback on the implementation of ethics policies.

8. Update Policies as Needed

The policies are there for a reason - and should adapt to changes in practice and law. At least one person in the committee is likely to need to keep abreast of new developments in AI and ethics.

9. Monitor and Review compliance

• Quarterly Reviews: Evaluate the effectiveness of the committee’s influence on AI initiatives. Is it working? Are decisions making a difference?
• Annual Audit: Conduct a thorough audit of AI ethics implementations and committee operations. Is the business running smoothly and in line with the principles first agreed upon.

As lawyers we're often the ones tasked with setting up these sorts of organisational structures. Anyone who has been through putting in place a GDPR governance framework will likely recognise many of the steps above.

*If it would help to have someone who has been through the process do let me know. Personally I'd suggest that we avoid the word "sub-committee" if that's OK by you? To me an "Ethics Board" sounds like a happier, and more decisive place. All those in favour?...