It’s possible that you could dial in some form of “cultural bias” with a knob. Let’s put it on the robot’s forehead just for fun. Depending on how far you turn the knob this would introduce a degree of conformance to laws, rules, and social norms, wishes of others, etc. Turn up the knob just a little and the robot will trade off low level risks in favor of conformance. Turn it up too far and the robot begins to ignore threats and active management of risks, and instead conforms to a generic scheme of orderliness that may or may not address actual risks associated with specific circumstances.

If you really want to know how I’d do it for a robot I would use much of the same software we used for Darpa’s “Urban Challenge”, the autonomous vehicle playoff. They actually DO follow all the rules of the road since you would be causing massive mayhem if you didn’t . But assuming I don’t give a fig about what happens to other road users, I might modify the software a bit – go ahead and cause other collisions if it would avoid me getting into one. But I have to admit I don’t see how I would at this point. The cars have collision detection systems that they already use to maneuver out of potential collisions. I suppose if I was programming for a skinny bicycle that also had really high robotic acceleration I might try to maneuver between two approaching cars, etc. as a last ditch avoidence method.

As for “programming” flesh and blood bicycle riding humans, I would recommend any fo the well-written articles and books on vehicular cycling.

Notice: I didn’t say anything about traffic laws. Man-made traffic laws can’t hold a candle to innate common sense.

I’m sorry, but this is the most utter nonsense I have ever seen. It is not based on any science or study that I know of but is apparently a random thought our of the bloggers head. Please don’t utter such things on a blog read by people trying to learn how to safely ride a bike.

The ideal way to bicycle is to learn an approach that minimizes conflicts with other vehicles. That way, if a bicyclist or motorist makes an error, as we humans do from time to time, the probability of a collision is low.

By minimizing conflicts, it also makes bicycling a more relaxed, enjoyable experience. That’s what vehicular bicycling is all about. It’s not about obeying some arbitrary set of rules. It’s about internalizing an approach to traffic to the point you don’t have to think about interactions most of the time.

It’s about making bicycling a safe, efficient and pleasurable activity, where you can spend your time enjoying the beauriful day from your bike, rather than spending every second dodging traffic.

His first rule of never slowing down or stopping for lights shows an incredible ignorance of what actually gets bicyclists injured or killed.

I hope noone took his advice and got themselves hit by a car.

Okay. Suppose that the robot were completely indifferent to traffic laws, but could be programmed with data on how motorists and pedestrians are likely to behave in various situations, including how they are likely to respond to a cyclist who uses hand signals, claims the lane, runs a light, etc. All this data could be included in the robots real-time risk assessment.

Would this robot obey, de-facto, traffic laws most of the time?

Are there situations where obeying the law increases exposure to risks?

As I said earlier (in a response to @rich), my commute contains a 50-yard lawbreaking run, which saves me time, decreases my overall exposure to traffic, and eliminates two left turns through intersections. I wonder what a risk-assessing robot would do?

My commute could be made 100% legal if I were willing to dismount and push the bike through this short stretch, but that would erase the time savings. Since this is my thought experiment, I’m going to say that the robot can’t dismount and push the bike.