The Absurd Dinner Bill
Since just about everyone reading this will have eaten at a restaurant at some point in their lives, we thought this would be a great example to explain the importance of AI Ethics when AI is in charge of your dinner check.
One night you go out to eat because you’re hungry and just can’t think of what you would cook at home. It’s been a long day, work was intense, and you’re looking to stop at your favorite restaurant where you know you’ll never be disappointed. Unfortunately, when you arrive, the restaurant has closed and has already been replaced by a new tech-infused restaurant named Your Eats, that can make any dish you want, even if you just show it a photo off of Instagram.
You don’t feel like searching for another place, and this is the only restaurant in sight. You enter, you scan your Citizen Score App on your phone, which identifies you to the establishment and verifies you have money to pay the bill. The new Citizen Score App the government implemented a few years back used to annoy you, but now it seems to make everything easier. You can even do your driving test for the DMV by syncing the Citizen Score app with your score. No more lines, no waiting. Get your driver’s license, take your exam or renew in a matter of moments.
You take a seat in what used to be your favorite spot in the previous restaurant and proceed to order. Taking your phone out and with the swipe and flick of your fingers, you open Instagram to show the desktop waiter a photo of your favorite dish from the old restaurant. The screen shows a total of $19.37, and you click accept.
The food arrives in a matter of minutes, which is way faster than the old place’s service. Sometimes you’d wait 30 or 45 minutes for your order. The meal is perfect, even better than you remembered it. The AI operating the restaurant must have read the comments you made about the meal somewhere and taken them into account. Now, this is how technology is supposed to work!
As you finish up your meal, the cameras throughout the restaurant continue to monitor your facial expressions and other biometric data, as they have been for your entire visit. The system prepares your bill and will send it as soon as you walk out the door. After gathering your things, you walk out of the restaurant, your phone dings with the bill, and you head to your car.
Settling into your car, you take out your phone to start the car’s engine, and that’s when you notice the most outrageous notification you have ever seen on your phone. The restaurant didn’t charge you $19.37; for some reason, the system charged you $307,802.01. The Citizen Score App had already emptied all of your liquid savings, checking, and money market accounts to partially cover the bill. The app was also set to sell all the positions in your retirement accounts, and the app has maxed out your credit cards. In the span of a few seconds, you had lost everything. What happened?
Rushing back to the restaurant, you look for someone to speak to, but there isn’t anyone human working there. Just screens. The other patrons at the restaurant notice the worried look on your face, and a nice young lady comes over to see if she can help. You quickly fill her in on the situation, and she goes about looking at the screen for a human operator button but no luck. She walks outside to see what the system charges her and when she returns, she mentions that she was charged less than what the system had quoted her when she ordered her meal. Other patrons started exiting the restaurant to see what they were being charged. Most were charged exactly what they were quoted, but some were charged only a few cents more or less than the quote they had received when ordering.
So what happened? Why were you the only customer to be overcharged so dramatically? Even if it’s an error, how can the system just take your money like that? Your life savings, your livelihood, your future? Well, this is where it gets interesting.
Interacting with the virtual customer service agent wasn’t much help. The only thing the system would say is, “Our system uses a variety of factors to determine the price of every meal to ensure the optimal experience of all patrons.” What does that mean? Sounds like a lot of very vague legal talk, where you say something but don’t really say anything. You figured you wouldn’t be able to get any answers that night since there weren’t any humans either in the store or elsewhere from what you could see.
The Next Morning
You spend the morning trying to get in contact with the corporate office of Your Eats, and after several hours, success! You got a real person on the phone. You explain your situation, try to crack some jokes about how absurd the situation is, but then things get even weirder than what you experienced the night before. Not only does the executive from Your Eats explain to you that after reviewing your situation, the system was correct to charge you over $300,000. In what world would a dinner cost over $300k?! You demand to know.
The executive proceeds to tell you that the system takes thousands of data features into account when assembling your bill, but since the system is so complex, they can’t provide any further insight.
You ask why you are being treated differently than all of the other patrons who at most had less than a one percent variance in their final bill. You are told that the system is fair and overall cases like yours are rare, but when they happen, there is no way to intervene and change the system from the company’s end.
You ask to see what are all the data features that are being used to calculate your dinner bill. The executive explains providing a comprehensive list of all the data they collected on you could harm the proprietary system and allow competitors to reverse engineer the system.
You ask if you can contest the actions of the system. You are told that given the complexity of the system and the way in which it is set up there is nothing that can be done.
Frustrated and feeling out of options, you ask who is responsible for the system and its outcomes? You are told that the system is responsible, not Your Eats, not an executive, not a developer. No one.
All you can say is that this isn’t how technology is supposed to work.
How this is applied to Ethical AI
If anything can be said about the story above is that it’s absurd in the extreme. But imagine for a moment that the system at Your Eats wasn’t calculating your bill incorrectly but instead was determining if you would be promoted, interviewed, admitted to college, sentenced to jail, granted parole, and so on. You might be rightfully concerned. It also might surprise you to learn that AI systems are already in use in these and other areas.
We are concerned with how we can keep the AI system fair for everyone in the world in a simple and straightforward way. We believe that all systems should be:
Transparent + Explainable
Improves Quality of Life
Accessible + Contestable
Incorporating these elements into the development, deployment, and operation of intelligent systems drastically reduces the possibilities and probabilities of undesired outcomes like the situation we described above. More importantly, when an undesired outcome occurs, the user would be able to understand what happened, contest the incorrect information, identify who is responsible, and advert any permanent consequences.
Now let’s dig into the rest of AI Ethics a bit more in-depth. Don’t worry, we won’t be getting too detailed. But you will have everything you need to understand how the Five Foundations of The Data Oath work, how they are applied, and what you can do going forward.
As always, if you have any questions, would like to get more detailed information, or would just like to share some thoughts with us, reach out here.