We humans are fickle. Our moods change all the time, but our AI-powered systems often have difficulty understanding this. How can AI become better at helping us discover what we want at any given time?
After a long day at work, you want to watch a show or a movie from a streaming service. Perhaps you usually like Nordic crime shows but are now4 up for something lighter. You may not even know what genre would be perfect right now. And neither does your AI-powered recommendation system. So instead of relaxing, you keep on scrolling aimlessly or resort to family and friends for recommendations.
These mood swings are where the recommendation systems can falter. The algorithm fails to understand that every occasion when a user searches for new content is one of a kind. Instead, it keeps on suggesting content that is similar to that consumed before.
“Somehow, I feel that the AI-powered recommendation system is a bit dumb. It’s like ‘If you watch a film with a spaceship, you must be interested in all kinds of space thingies.”
Male, 28, Helsinki
And then there’s the situation when you get into something new. Perhaps you heard a really good song in a movie and got interested in a genre of music you haven’t listened to before.
In the beginning, the AI-powered system may give the user good suggestions as they develop a general understanding of their new interest. Soon, the journey of discovery comes to a halt, as the system keeps on recommending only songs similar to the first ones and doesn’t let them explore further.
The AI-powered systems are often not sensitive enough to people’s developing tastes, so they can’t support the user on the journey of discovery. The user becomes frustrated and resorts again to external sources for guidance, as there is no way to collaborate with the system.
In both of these cases, the challenge arises from the fact that while humans, their moods, and tastes change constantly, the algorithmic systems remain reactive: they base their recommendations on what has previously happened, not what users would want to happen next.
Sometimes people want to take a passive role towards AI (link to post #1), but sometimes they want to engage in algorithmic decision-making. To better accommodate people’s current moods, there should be more effective and convenient ways to poke the algorithm in the right direction. It can be as simple as a way to let the system know how long the recommended content should be.
And if the user feels that AI isn’t helping them on their journey, there should be a way to inform the system of the user’s evolving interests and see the algorithm react to that. Sometimes we all need to break free from our user profile to discover something we don’t know yet.
Everyday AI is a collaboration between Alice Labs and the Centre for Consumer Society Research, University of Helsinki, in partnership with Reaktor. The Engaging With EverydAI webinar took place on 5th May at 9am CET / 10am EET.