How to Build and AI Governance Program

I joined an artificial intelligence company because I had the “if you can’t beat them, join them” mentality. I knew A.I. was the technology wave of the future, so if I wasn’t going to be the one to stop it, then I needed to be a part of building it. These types of companies in particular need someone whose job includes building an ethics strategy that fits into the overall business strategy. I was eager for that challenge. By the time I left, I had a plan and I am giving it all to you right here, right now.

Let’s face it, A.I. singularity is a scary concept. So, if you don’t have someone who is focusing on ethics, then you might as well stop reading this right now and go out and hire someone — an executive someone that we will call the lowercase ceo, “chief ethics officer.” See you back here when that step is done. Hmmm, I see potential for another story on the ideal job description…

Okay, with that out of the way, here is how I recommend A.I. companies build an ethics strategy. Remember, this is just my personal opinion so all the usual qualifiers apply (I am a lawyer after all).

First, building an ethics program has to start with your company’s core values and then mapping ethical principles to them. Sound hard? Well, it is.

During my early days at Clarifai, I wrote a blog post on A.I. and ethics (since taken down). I suggested some ethics principles at the end of that post that I had gleaned from academia and law. Lots of professions are guided by ethics, so definitely do your research. Or you can just take from my blog post and match them up to your core values. I don’t actually recommend relying on my blog post, but it’s there and I did do the work. Anyway, let’s say one of your company’s core values is responsibility. Responsibility could be matched to “legality” in ethics, i.e. be fair and just, or it could be matched to “honesty,” i.e. to be transparent. Both are great but don’t make the mistake of incorporating both. You have to choose one. More on that in a second, but, please, root your decisions in the company’s culture. Why? Because if you don’t root it in your company’s culture, then you will be fighting a very steep uphill battle.

If you are the newly hired ceo, make sure you understand the company culture before this first step. I am crossing my fingers for you that the culture isn’t unworthy and in need of a change because that’s a whole other post.

Second, now you have your set of ethical principles. I believe in the power of three’s so hopefully you did not set too many. In this case, less is more memorable and the ethics for your company have to be memorable by every single employee. Don’t go crazy.

This brings me to an important point. You are not responsible for making sure the company is behaving ethically because no one person should become the ethical police officer. Besides, that turns into either laziness or a usable excuse by the rest of the executive team. Frankly, it’s impossible anyway because one person cannot be everywhere. I could write a whole post on this one point.

So, the second step is to build your team. I believe this should be a committee, actually, an official Board committee. The Board committee will need its own charter and the independent Board member should sit on it, if you have one. If you don’t have an independent Board member or enough Board members overall, then form an Advisory Board and appoint one of them to what should now be called the Advisory Board committee (same function, with a little less formality). Whichever one you go with, you should present on behalf of the ethics committee (no less than 1 time per year and ideally quarterly) to the full Board of Directors. Of course, the members of the ethics committee should meet more regularly, outside of Board meetings.

Finally, you need at least, at least!, one more person on the Advisory or Board ethics committee (we will call this the “BEC” from now on) whose job is either partially or fully focused on applications where your company’s A.I. is being used for “the greater good.” Ideally, this person is skilled in building your company’s product and is interested in ethics. You can include other people on the BEC, like a product person would be great or a customer success director or a data scientist. But don’t let it get too big. If you are a huge company, then just have multiple sub BECs that report to the one big BEC.

One small aside, please, please make sure your committee is as diverse as possible. Diversity and inclusion is the foundation of any AI ethics strategy (this cannot be emphasized enough and, likely, will be another post).

Third, now the BEC has its guiding ethical principles and a team to evangelize them both internally and externally through word of mouth as well as use of the product for good. Right on.

You didn’t think that was all it was going to do, did you? Of course not. Now that you have been promoted to the ceo, you have to use your position in the company to take a broad view of all business units and what they are up to in case the ethics principles need to adapt. You also need to be on top of industry trends in ethics, maybe even joining a network (I am looking at you Partnership on AI) or other industry groups that like to discuss ethics in A.I. And, of course, you should be reading - a lot. You are also responsible for calendaring the BEC meetings, setting the agenda and planning. ABP — Always Be Planning!

The other non executive, non Board members of the BEC are responsible for implementation and monitoring of the ethics principles throughout the organization. If you can find a way to measure and track data about how effective the ethical principles are being adopted, through general observation or maybe even surveys, and bring this back to the BEC meetings, that would be ideal.

Fourth and finally, the BEC should use each meeting as part working group session to determine and track key metrics that they can use to report to the Board of Directors (another one worthy of a separate post). They need to empower the Board of Directors with how the organization has positioned itself on ethics with use cases or antidotes the Board of Directors can mention to the public. The CEO, in particular, needs to be regularly briefed on the BEC’s work so that the CEO can speak to it intelligently and confidently.

Building an ethics strategy and program into A.I. companies is not easy, but it is time and money well spent and I hope that you, the employees or the investors, will take the need really seriously.

Recap: So remember, to build an ethics strategy and program you will need to:

1. Hire an executive whose role is, at least partially, dedicated to running the ethics strategy. Chief Ethics Officer anyone?
2. Match your company’s core values to a set of ethical principles.
3. Diversely staff the committee and formalize it.
4. Hold regular meetings.
5. Track and report the BEC’s work to the Board of Directors.

When it comes to A.I., advancement is inevitable. BUT, remember, just because you can build it, doesn’t mean you should. But, then, who am I to judge. :)