Creating AI goes beyond the algorithms, and may be more about the humans that use it.
At CAIDEL, we spend a lot of time thinking about how to make AI as accessible as possible. We have noticed that just having a good algorithm isn’t enough. Taking it literally, you can say that studying psychology is a must to get AI successfully up and running. That’s why attending “Beyond Technology: Utrecht AI Event” was especially interesting to us.
“Human-first AI.” I feel that’s the main takeaway from the day. The event brought professionals from various fields together. From ethics, healthcare, security, and psychology, central was the impact of AI on human life and what it takes to make AI successful beyond its technology. “Does AI take away the need for human work?" "Does AI make me lazy?” “Can I really trust AI?" Some of the talks were specifically about use cases, such as in healthcare, which, beyond their industries, still gave interesting insights we all can use.
Making AI “human-first” sounds straightforward at first. After all, the ones using the technology are humans. But the talks showed that it goes beyond just making lives easier using AI. In fact, “using AI” should ideally be removed from the user’s perspective. Integration in existing workflows gives a much better adoption rate compared to introducing an AI tool standalone.
There also exist differences from the point of view of the scientists, creating the AI, versus the users who’ll be using the eventual product. UI placements that I as a developer find logical might feel strange to users outside the field. Some of the talks went deeper into this, with one specifically on human behavior by Professor Judith Masthoff. Another I attended was a panel led by Prof. Dr. Karin Jongsma, going into use cases such as real-world insights on how people in healthcare react to AI. The latter had a funny anecdote. Clinicians evaluated an AI tool and voted “usability” (instead of more common aspects like “human in the loop”) as the most important aspect of AI use. This, while usability isn’t much mentioned in upcoming AI regulations at all.
It kind of makes sense. The easier to use, the easier to adopt, and the easier to prevent mistakes. More of the headspace can be used to evaluate the result, instead of figuring out how to get AI to come up with the result.
For more on human behavior, check out the blog post I made on it inspired by the event.
The more AI becomes integrated in our lives, the more important it is to assess its risks. Questions around bias, fairness, and unintended consequences need to be addressed early and continuously. Showing the why alongside the what helps users understand what the AI uses as a basis of its claims, something that can later on be verified manually if needed. This is especially important in healthcare, and it’s nice to see that it’s taken into account by the AI developers too. Beyond critical fields, from the talks attended I can see it’s also important to users, giving them trust in the response and control over what they let AI do.
Furthermore, AI doesn’t need to be used to give the perfect result. In healthcare, I’ve seen some examples in the talks that show AI used to generate drafts and noncritical judgements. This way the end-user stays in control, lowering the risk of AI by change of use instead of the underlying technology.
We stood alongside Engineering Trainer, who’s currently running our pilot program on their website. Engineering Trainer provides an online platform for engineers to master new skills used in real projects. CAIDEL helps them with conversion rate optimization (CRO), by upgrading the search experience with course recommendations through natural language search. We make sure to give fair, accurate results that Engineering Trainer can inspect themselves using our client app and the evaluation reports we make for them.
Check them out if you’re interested, and check us out if this sounds interesting for your business!
I can confidently recommend Beyond Technology and other AI events that go beyond the algorithmic aspects. It was refreshing to see insights from others and other fields on something I actively work on at CAIDEL. Staying in touch outside your own bubble is important to know what impact you will bring with your emerging technologies.
There were many other enlightening insights I gained during the Utrecht AI Event. All presentation slides are available online, which I happily advise going over to see how Utrecht is booming with AI innovations, small and big. It really inspires me what we can do as a society.
I’m sure next year will be just as insightful and fun!