Follow Us on Google
Oladimeji’s path from marketing executive to artificial intelligence (AI) researcher wasn’t planned, but it was inevitable.
As Regional Marketing Manager for Intake Education across Africa, she managed £560,000 in marketing budgets, led campaigns for over 100 universities, and helped thousands of students navigate international education.
But the technology meant to help often felt like a barrier. As a doctoral researcher in Information Systems at Louisiana State University, she is now tackling one of the most critical problems in tech today: AI systems that make decisions people can’t understand and therefore can’t trust.
In this interview, Mathilda explains why explainability matters, how her marketing background influences her research, and more importantly, the things that need to be changed before AI can truly serve everyday users.
You went from managing university marketing campaigns across Africa to researching AI explainability. What connected those experiences?
Curiosity, honestly. At Intake Education, I ran digital campaigns for over 100 universities across Nigeria, Ghana, Kenya, and other markets. I could see how students were using platforms powered by algorithms that decided which universities they’d see, and which options got prioritized. Now it’s undisputed that all these were life-changing decisions, where to study, how much debt to take on, what career paths to pursue; all carry serious, long-term consequences.
However, nobody could fully explain why the system showed what it showed. I could see the backend data, but the students? They were just trusting a black box. That gap gnawed at me. I didn’t want to work around it. I wanted to solve it.
So the trust problem with AI isn’t abstract for you. You saw it affect real people.
Exactly. And it wasn’t just students. Even my marketing team struggled. We’d get recommendations from Google Analytics or Facebook’s algorithms, and trained marketers couldn’t explain why the system suggested what it did. We were all just hoping it was right. If people with data literacy were confused, what about everyone else? That’s when I realized this wasn’t just a marketing problem. It was a technology problem.
Your research at LSU focuses on AI explainability and user trust. What does that mean in practice?
It means understanding how AI systems can reveal their reasoning, not just deliver answers. Most AI operates like a black box, input, output, no visibility into what happens in between. For anything affecting people’s lives, healthcare, loans, jobs, education, that’s unacceptable. My research examines what makes an explanation actually useful.
Because you can technically explain how AI works, but if it’s full of jargon only experts can understand, you haven’t solved anything. True explainability means making systems understandable to the people using them, not just the people building them.
You’ve published at ICIS and AMCIS on this topic. What have you found?
What we’re doing is examining how AI explainability affects user trust and even decision-making. When we have systems that are opaque, people face constant uncertainty, they won’t understand what they’re agreeing to or why they’re getting certain recommendations.
Other News
So in the end, they either blindly trust the system or reject it entirely. Neither is sustainable.
We’re also synthesizing explainability research, mapping different approaches and separating what genuinely improves user understanding from what merely sounds sophisticated.
Very well, so what is that thing that needs to change in how the tech industry approaches AI?
That is the idea of treating explainability as an afterthought.
Many companies build AI first, afterwards, they then try to make it explainable only when regulators demand it.
That’s not appropriate, instead it’s backwards. Explainability needs to be built into the design from day one.
Even that is not enough. We also need to democratize AI literacy. Not everyone will become an expert in the technicalities, but everyone may eventually be affected by AI decisions. Everyone, whether tech savvy or not deserves to understand the basics of how these systems work.
That’s as much an education challenge as a technical one.
So if you could reform one element of AI development, what would it be?
That will be to establish and enforce mandatory transparency standards. At the moment, there are no universal requirements for AI systems to clearly explain their decisions in accessible terms.
Big companies deploy algorithms that affect millions of people without ever showing their work and they face no consequences for this.
At this point, what we need are regulatory frameworks that make explainability a very crucial requirement, not just a nice-to-have feature.
When decisions like this affect people’s lives, then they have a right to understand how those decisions are made.
Lastly, what does success look like for your research?
It’s simple and direct: AI that feels less like a black box and more like a relatable tool people actually understand and control.
The role of technology should be to expand access and opportunity, not gatekeep it. This is exactly what led me into marketing, helping students to get access to education they thought was out of reach. It’s the same drive now.
If my research contributes to AI systems that everyday users can understand and trust, systems that serve people instead of mystifying them, that is success.
