Calgary spent taxpayer dollars letting an algorithm spit out fifty intersections where pedestrians are apparently most likely to get flattened. Officials patted themselves on the back, issued a press release about smart city innovation, and pretended they actually solved something. It is a comforting bureaucratic bedtime story. It makes the city council look proactive while changing precisely zero equations on asphalt.
The lazy consensus in modern urban planning says that if you feed enough historical crash data into machine learning models, the computer will whisper the secrets of traffic safety into your ear. This is a delusion. Algorithms do not predict human stupidity or municipal design failures. They merely memorialize where blood has already been spilled.
I have spent two decades tearing apart traffic engineering reports, municipal budgets, and collision metrics across North American transit corridors. I have seen cities blow millions on reactive dashboards that tell them what they already know: cars hit people where wide lanes, high speed limits, and poor lighting invite drivers to accelerate.
Let us look past the marketing gloss of predictive analytics.
The Flawed Premise of Predictive Black Spots
The entire premise of identifying high-risk intersections through algorithmic modeling rests on a statistical mirage. When a system flags fifty dangerous spots, it is fundamentally treating the symptom while ignoring the pathology of the grid.
Traffic engineers love to blame driver distraction or pedestrian jaywalking because it shifts liability away from the geometry of the road. If a street is built like a runway, drivers will treat it like a runway. No neural network can compensate for a five-lane stroad designed to move vehicles at highway speeds through dense commercial zones.
When Calgary feeds past collision data into an AI, the software outputs a list of locations where the probability of a crash has historically been highest. This is not foresight. This is archaeology. You are paying for a digital Ouija board that reads the tea leaves of past tragedies.
What happens after the city flags these fifty intersections? They add a fresh coat of thermoplastic crosswalk paint, maybe install a flashing LED sign that drivers learn to ignore within a week, and check a box on a quarterly performance review. The underlying velocity of the corridor remains untouched. The design remains hostile.
Why the Algorithm Is Looking in the Wrong Place
Ask any veteran transportation safety auditor what causes severe pedestrian impacts, and they will point to two variables: kinetic energy and friction points.
Kinetic energy is a simple function of mass and velocity. Hit a pedestrian at thirty kilometers per hour, and they have a ninety percent survival rate. Hit them at fifty, and that survival rate plummets below twenty percent. Algorithms can rank intersections from one to fifty all day long, but if the default speed limit through those corridors stays at fifty or sixty kilometers per hour, the software is an accomplice to the physics of destruction.
Friction points are where turning vehicles cross pedestrian paths without physical protection. Traditional intersection design prioritizes vehicle throughput over human life. Permissive left-hand turns—where drivers must scan for pedestrians while gap-seeking against oncoming traffic—are slaughterhouses. An AI does not need to analyze five years of historical data to tell you that mixing heavy turning vehicles with unprotected foot traffic in a poorly lit intersection will produce casualties.
Yet cities cling to software solutions because hardware solutions require political courage. Narrowing a lane, removing a turning pocket, or choking down traffic capacity upsets commuters. It creates gridlock complaints. So instead of fundamentally altering the geometry of the street, administrations buy algorithms that give them permission to tinker at the margins.
The Counter-Intuitive Truth About Safety
Real safety is not about predicting where crashes will happen. It is about making it physically impossible for drivers to achieve lethal speeds in spaces shared by human beings.
If you want to protect pedestrians in Calgary, turn off the computers, throw away the heat maps, and implement aggressive traffic calming across every single neighborhood arterial.
Here is what actually works, stripped of consultant jargon and tech utopianism:
- Widen sidewalk bulb-outs at every urban intersection to shorten crossing distances and force motorists to take tighter, slower turns.
- Eliminate permissive green phases for turning vehicles wherever foot traffic crosses. Give pedestrians an exclusive lead interval or ban right turns on red universally.
- Drop absolute design speeds by physically altering road widths. If a lane is wider than three meters, you are inviting disaster, no matter how many warning signs you post.
- Remove mid-block crosswalks on multi-lane undivided roads. They are sacrifice zones. If pedestrians need to cross four lanes of traffic, build a raised median refuge island or a grade-separated crossing. Anything else is Russian roulette with a crosswalk button.
The contrarian reality is that smart city initiatives often function as digital smoke screens. They project an image of hyper-competence while preserving the status quo of car-centric infrastructure.
Calgary does not need a better algorithm to tell it where pedestrians are dying. It needs the spine to admit that the way it builds streets is fundamentally broken. Stop outsourcing accountability to code. Stop treating municipal safety as a data science problem when it is an architectural moral failure.
Pull the plug on the predictive models, pick up a shovel, and start narrowing the roads.
