Fixing bikes and Explaining AI; The Unexpected Parallels

Bike Kitchens, at their core, are about empowerment. They teach people to understand their bikes, work on them, and regain confidence in their ability to move through the world. As I spent more time fixing bikes and teaching others, I began to notice surprising parallels with my academic work in explainable AI (XAI).

AI models, much like bikes, can sometimes seem like black boxes to those who use them. A cyclist might not understand why their chain keeps falling off, just as a data scientist might struggle to explain why an AI model is misclassifying certain inputs. In both cases, it’s about breaking down the problem into understandable parts, diagnosing the issue, and applying a fix.

Getting into Bikes

I got into bikes out of necessity. After moving to Amsterdam for my master’s, I had the rather typical experience of needing a bike. A quick search on Marktplaats led me to De Pijp, where I met the Professor. The Professor was a skinny Dutch man with scarce dangled white hair and a face tired and wrinkled by years of many rainy winters. He greeted me at the door by rolling a cigarette and invited me in to present an old green Dutch bike. It wasn’t your typical Dutch city bike frame but hand-built with six gears, and he threw in a chain with it, too. For all 70 Euros, I was very happy.

Until about 10 minutes later, when the chain snapped at the end of the corner. I walked back to ask for help, but instead, I got my first lesson from the Professor. He got a chain tool and showed me how to work it. My hands got greasy, but the chain links popped out, and I got home on my new bike just fine. But my tribulations hadn’t ended, just begun. That’s how I discovered the Bike Kitchen at the UvA.

What are Bike Kitchens

The Bike Kitchens are workshops run by volunteers, activists, and socially engaged cyclists. They are convivial places to repair bikes, teach hands-on self-repair and maintenance skills, and give free bikes to those needing them. l’Heureux Cyclage operates a network of 250 workshops throughout France. In Belgium Cycloperativa operates 18 arrondissements around Brussels, among other Bike Sheds around the world.

These places aim to promote bicycles as a sustainable form of travel by offering low-cost second-hand bikes and repairs. Bike Kitchens are also anticapitalists, aiming to refurbish bikes, fix them up, and minimise abandoned bikes. This way, Bike Kitchens are creating a circular economy of bikes, thereby engaging many new people in sustainable initiatives.

Bike Kitchens also offer a space for knowledge-sharing. Many cyclist experiencing issues with their bikes either ignore it, putting them and others at risk and often worsening the issue. Or they turn to expensive mechanics and, upon hearing the repair cost, abandon the bicycle altogether. Basic mechanical skills required for bike repair are extremely versatile and useful in many areas of life, so in bike co-ops, you can learn to maintain your bike, diagnose problems and solve them.

Diagnosing Problems: Bikes vs. AI

When a bike stops working, the first step is to figure out what’s wrong. Is the chain slipping because it’s too loose? Are the brakes squeaking because of worn pads? Similarly, in XAI, understanding why a model made a certain prediction involves identifying the functions and workings of the model. Tools like SHAP values, counterfactual explanations, or mechanistic interpretability help break down a model’s decisions into interpretable components.

Without the right tools or guidance, the diagnosis process can feel daunting for both bikes and AI. That’s where spaces like Bike Kitchens and XAI research play a crucial role—they demystify complex systems and make them approachable.

Empowerment Through Understanding

But Bike Kitchens don’t just fix bikes—they teach people to fix their own. There’s an incredible sense of empowerment in learning how to true a wheel or adjust a derailleur. Similarly, XAI seeks to empower AI users by giving them the tools to understand and challenge model decisions. Whether it’s a doctor interpreting an AI-assisted diagnosis or a policy analyst evaluating a model’s fairness, explainability ensures that AI is not just a tool but a collaborative partner.

Much like AI training is an iterative process, Bike Kitchens promotes sustainability through repair and reuse, breathing new life into what others might consider junk. Instead of discarding models when they fail, researchers iterate, retrain, and refine them. The process is about learning from past failures, making improvements, and building something better. Sometimes, mixing and matching components of pipelines and polishing individual parts make larger breakthroughs possible.

Perhaps the most striking parallel is the emphasis on community. At a Bike Kitchen, the focus isn’t just on individual repairs and fostering a supportive space where knowledge is freely shared. Similarly, the open-source AI community thrives on collaboration, with platforms like Hugging Face enabling researchers to share models, tools, and insights.

By working together, we create a culture of mutual aid and shared progress—whether fixing bikes or building explainable AI systems. It’s not just about solving immediate problems but about creating systems that are fairer, more sustainable, and accessible to all.

Conclusion

While bikes are much simpler systems than modern large language models and other complex AIs, both are embedded in our lives and systems. To create safe, sustainable and trustworthy AI systems, we might take some inspiration from the humble bicycle.

Through Bike Kitchens, I’ve learned that fixing something is never just about the object—it’s about building confidence, sharing knowledge, and fostering a sense of ownership. In my work as an explainability researcher, I aim to bring the same ethos: demystifying complex systems and empowering others to engage with them meaningfully.

Whether it’s repairing a bike or explaining a neural network, the goal is the same: to help people move through the world with a little more clarity and confidence. And sometimes, a little grease on your hands.