Developments of artificial intelligence are a source of concern for several reasons. First of all AI threatens jobs, privacy and our perceived freedom. Second, 99,99% of the population doesn’t really understand what AI is and where it originates.
In today’s post I’d like to raise the case of respectful Artificial Intelligence and explain why we need the way we conceive “intelligent” algorithms.
Why do we need respectful artificial intelligence and privacy laws ?
Technology should be at the service of Humans. Not the opposite. This idea is pivotal in Cathy O’Neil’s book “Weapons of Math Destruction“. Like any other tool invented by Men, it needs to be harnessed to reveal its full potential. Yet, as O’Neil explains in her book, AI tends to be mainly used to enslave other men and to increase inequalities. Why is that ? Because there is a fortune to make in AI and that’s why 6 software giants teamed up to defend their viewpoint and get general consent on their vision of data collection and AI uses.
We need respectful algorithms
While efforts can be deployed to be more transparent towards users in terms of privacy, creating respectful algorithms is in my opinion a much more complicated task. Avoiding the creation of Weapons of Math Destruction (WMD) requires that the very designers of those WMD work not only at increasing value for the firm but also for the users. An even greater challenge is to be allowed to do so; in other words a firm should be convinced that respecting the users, focusing on the value he or she extracts from using algorithms, is more important in the long-run than focusing on the value extracted by the firm.
Isn’t it extraordinary that the basic principle “satisfaction creates loyalty” seems to be more and more forgotten ?
Loyalty by design, not by choice
Artificial intelligence has brought a revolution in terms of marketing. The dream of a personalized 1-to-1 marketing is about to become true.
Algorithms have made it possible to adapt to our predicted behaviors. Improving loyalty has become a machine-learned mechanism, a feature that pushes the consumer to repeat purchases. But is this real loyalty ? In my opinion it’s not. Real loyalty comes from the desire to be loyal to a brand. I can remain loyal to a brand, have strong positive feelings towards it, without necessarily purchasing compulsively.
On the opposite I can purchase from a brand without necessarily being loyal to it. Take Uber for instance. I happen to purchase from that brand but don’t have very strong feelings towards it. Actually if an equivalent service was available, I’d prefer to switch to it provided drivers are treated correctly and that taxes are not evaded to fiscal paradise.
How to develop respectful algorithms ?
In a previous article I argued that we needed to “open up the models“. Black boxed algorithms are dangerous because control on how results are produced is difficult, even impossible. That’s how WMD are created : through insufficient control possibilities.
Not only do future respectful algorithms need to be more transparent but they need also to be the product of shared wisdom. My argument is that algorithms 2.0 will need to be co-created. Like BNP Paribas co-created its privacy policy, tomorrow’s Data Science leaders will increasingly turn to customers to co-create algorithms. If consent must be gained from users to collect and use their data, wouldn’t it be all the most logical to associate users in the very definition of what is proposed to them and how it is proposed ?
Final words
Time has come to break the pattern of unilateral decision making. It is true in a certain sense that consumers don’t know entirely what they need. I won’t repeat Henry Ford’s famous saying.
Yet, even if consumers can’t imagine the world of tomorrow, does it mean that those innovations can be created without them by people who have little knowledge of how their inventions will affect the lives of those who will use these inventions ?
Posted in big data, Innovation, Marketing.
