How Can We Trust AI Companies?
When building a company or a product, a lot of times you end up at a crossroads where your integrity and commitment to being a company people can trust is tested. If you’re in the medical or banking fields, you don’t even have a choice — there are regulations instead. Not all of them are perfect, but it’s better to have some rather than none.
When building a commercial entity, however, you have the full spectrum at your disposal. If you’re in the Artificial Intelligence business, and you work with people’s private data, you have a huge responsibility.
Defining Trust
In his post, The Cost Of Our Dishonesty, Dan Ariely explains how easy it is to lose trust. When trust is lost, regaining it is one of the most difficult challenges in a relationship — it might be the hardest of all.
A lot of people might think that a trustworthy entity or person is necessarily a nice and comforting one. Good = trustworthy, bad = not trustworthy. This assumption couldn’t be more wrong.
Here’s an example. On one hand, there’s my bank — I store my money in it, I make almost daily transactions using it. Do I trust it not to withdraw account fees which I wasn’t fully aware of? And would I expect it to admit its mistake later on and fix it? Unfortunately, no. I wish I could. On the other hand, when Anonymous says they are going after an entity, I somehow assume that they will live up to their promise.
So what is trust all about? And how can a company which deals with people’s personal data, where limits are still vague, make sure it’s worth the human trust?
Intentions
Why are you in this business in the first place? If your business model revolves around selling users’ data, you don’t have a horse in this race. Please stop reading this post. Sorry for wasting your time.
Are you trying to bring a better value to your customers? To make a change? To challenge existing old concepts in your field? Why are you doing what you’re doing? Not to be mistaken, of course you want to grow a successful business and earn money. But you need something to stand for and live up to. A person without a vision is like a ship without a compass.
Managing Expectations
This is where people test your intentions vs. real life. A promise delivered as expected earns you trust points. If you under-deliver — you lose some. So far, so good?
Now here’s the interesting part: many would think that over-delivering necessarily leads to a positive experience. The reality is that even though, as a one-time occasion, it can be pleasantly surprising, if it keeps happening all the time, it can actually reduce trust. Imagine an employee who constantly estimates tasks taking 2x the time it actually takes him to get them done. Would you trust his next estimation? Now imagine a company doing that. What would you think of its prognosis abilities as a board member or a stakeholder?
Transparency
You don’t have to bore your customers with information they don’t need, but if there’s something they should know, and you’re deliberately not telling them, then welcome to the untrustworthy circle.
Keeping an open channel with your customers is crucial to communicate your intentions, manage expectations, and keep everybody up-to-date in case something goes extremely well or extremely wrong. We all fail many times before we reach a good result. Communicating the process will make your customers feel fully engaged with it. People are more likely to trust someone who makes mistakes and is open about it because they trust them to fix these mistakes.
Control
A person who consumes a product which uses AI over their private data should be able to know what data is used (transparency) and what to expect out of it (managing expectations). A person also has to be able to make a choice to opt-out if they wish to. Even if used by 0.0001% of your customers, giving customers that sense of control will make them feel comfortable and trust you more.
Trust Costs You Money
When looking at the short-term, you can probably make a whole lot of money by secretly gathering users’ data in your products like Samsung and Bose did. But it’s an overhead. It will always bite you in the ass eventually and might actually make you lose that exact money you worked so hard to earn.
AI and Trust
For service providers in AI and personalization fields, it’s an amazing opportunity to understand what happens in a person’s life. We, people, are curious creatures in general, and data scientists are craving for data to make products greater. However, AI has and will have frightening capabilities, and it’s a big deal who runs the show.
So, what kind of company are you building? Can it be trusted?
