Road Safety Is Not Just Potholes

Recent advances in computer vision have enabled companies to detect pavement defects—potholes and cracks—with high accuracy. These technologies are being marketed as road safety solutions. Yet simultaneously, highways designed for 60 km/h but traveled at 100 km/h, lacking median barriers and shoulder protection, continue to cause fatalities.

This disconnect highlights a fundamental misalignment between what road safety technology is addressing and what actually causes crashes and injuries.

The Easy Problem vs. The Hard Problem

Let me explain the dynamics at work. Potholes and pavement cracks are readily detectable through automated image analysis. You can apply a CNN to detect defects. The task is well-defined, the problem is bounded, and success is measurable through standard metrics.

Road safety infrastructure assessment, by contrast, is inherently complex. It requires understanding how median barriers affect crash severity. It requires evaluating whether guardrails are installed correctly. It requires assessing whether pedestrian crossings are positioned safely. It requires determining if speeds are appropriate for road design. These are more challenging problems. They require domain expertise in road safety engineering. They involve multiple interrelated variables.

This creates an incentive structure where technologies address the more tractable problems—pavement defect detection—rather than the more complex ones.

What Actually Kills People

In India, where roughly 172,000 people die on roads annually, the leading crash factors are not potholes. They are:

• Speed inappropriate for road design: A 100 km/h highway without median protection, where impacts are fatal instead of survivable
• Missing or inadequate barriers: No guardrails on curves, inadequate median barriers on divided highways
• Unsafe pedestrian infrastructure: No footpaths, no safe crossing points, pedestrians forced into traffic
• Dangerous intersection design: Poor sight lines, conflicting traffic flows, inadequate signal timing
• Lack of street lighting: Nighttime crashes where drivers can't see hazards or pedestrians
• Vulnerable road user exposure: High speeds near schools, markets, residential areas

Evidence from crash data indicates that pavement quality contributes minimally to crash causation. While a pothole might cause vehicle instability, the primary factors determining crash severity remain infrastructure design features. A highway without a median barrier operating at 100 km/h presents significant risk for head-on collisions. A pedestrian crossing in the middle of a multi-lane highway without a refuge island faces substantial danger.

The Real Standards

Road safety standards evaluate roads on attributes like median barrier type, presence of edge line markings, roadside hazards, pedestrian facility quality, and posted speed appropriateness. These attributes, when present and properly implemented, reduce crash severity by 50-80%. A road can have perfect pavement and still be unsafe if it lacks these features.

Conversely, a road can have minor pavement defects and be relatively safe if it has proper median barriers, appropriate speeds, and pedestrian protection.

When a product is marketed as "AI for Road Safety" but only detects potholes, it is addressing what is technically tractable rather than what is most consequential for safety.

The Distinction Between Maintenance and Safety

There is a meaningful distinction between road maintenance and road safety that is sometimes conflated in technology marketing. Road maintenance—including pavement defect detection—contributes to infrastructure longevity, ride quality, and vehicle performance. These are legitimate engineering objectives.

Road safety, however, addresses crash prevention and injury mitigation. These are distinct objectives requiring different approaches. When a company presents pavement defect detection as a road safety solution, it is conflating maintenance improvement with crash prevention.

This conflation may be inadvertent or intentional, but it results in misaligned resource allocation. Budgets designated for safety improvements may be directed toward maintenance technologies, while actual infrastructure safety deficiencies—missing barriers, inadequate pedestrian facilities, inappropriate speeds—remain unaddressed.

What We Should Be Detecting Instead

If you want to actually improve road safety with AI, detect:

• Missing median barriers on high-speed divided highways
• Inadequate guardrails on curves and embankments
• Absent edge line markings on rural roads
• Missing or inadequate pedestrian crossing facilities
• Presence of roadside hazards (rigid poles, trees too close to road edge)
• Speed inappropriateness for road design
• Inadequate street lighting for nighttime visibility
• Lack of footpaths or cycle lanes in mixed-traffic areas

These are the attributes that determine whether a road is safe or dangerous. These are harder to detect. They require understanding road design principles, not just image classification. They require domain expertise. But they actually save lives.

Resource Allocation Implications

Government budgets and investment decisions are informed by what technology vendors present as road safety solutions. If pavement defect detection is positioned as a road safety intervention, budget allocation may prioritize this work over infrastructure safety assessment. The result is that assessments of critical safety attributes—barrier types, pedestrian facilities, speed appropriateness—may be delayed or unfunded.

Meanwhile, the underlying infrastructure deficiencies persist. Median barriers remain absent. Pedestrian crossing facilities are inadequate. Speed limits remain mismatched to road design.

Toward More Comprehensive Approaches

Companies developing AI for road safety should clearly specify what problem they are addressing. If the technology detects pavement defects, this should be described as a maintenance and condition monitoring tool—valuable for asset management, but distinct from crash prevention.

True road safety automation requires evaluating infrastructure design attributes: barrier types and placement, edge line markings, roadside hazards, pedestrian facility quality, and speed appropriateness. These assessments demand domain expertise in crash mechanics and road design principles. They are more complex than pavement defect detection, but they directly address factors that determine crash outcomes.

Comprehensive road safety assessment integrates multiple types of analysis: infrastructure design evaluation, crash data analysis, speed management, and vulnerability assessment. Pavement monitoring is one component of comprehensive asset management, but it should not be confused with road safety assessment.

Evaluating Road Safety Solutions

When evaluating AI solutions for road safety, consider: Does the system assess infrastructure design attributes identified in road safety standards? Can it identify missing or inadequate median barriers, guardrails, or pedestrian crossing facilities? Does it evaluate speed appropriateness for road context?

Solutions that focus exclusively on pavement condition should be appropriately scoped as infrastructure maintenance tools rather than road safety systems, even though good maintenance contributes to overall infrastructure performance.