When Safety Feels Urgent: Bias, Evidence, and Decision-Making in Agility - The Dogwalk

A response to current questions about dogwalk safety.

Photo courtesy of Blue Merle Photography HDG

Written by: Dr. Kim Cullen, Associate Professor, School of Human Kinetics and Recreation, Division of Population Health and Applied Health Sciences, Faculty of Medicine, Memorial University of Newfoundland

Witnessing a serious injury happen in sport is always upsetting. At the recent Winter Olympics (Milano-Cortina 2026), for example, rare but dramatic injuries in events like alpine downhill skiing, short-track speed skating, and ice hockey captured global attention. When injuries are uncommon but highly visible, they tend to dominate discussion and evoke a strong desire to act.

That urgency comes from a good place. But it also creates conditions where decisions are more likely to be driven by emotion, gut instincts, and incomplete information. One lesson we can take from high-hazard industries, where mistakes can lead to serious injury or death, is the role of accident investigations: not to assign blame or act immediately, but to slow the process down long enough to understand what happened, why it happened, and whether proposed changes would actually reduce risk rather than simply respond to a single outcome.

Understanding why something happened is the first step toward responding responsibly.

These debates are hard because urgency lets bias take the lead

Safety debates around rare events are difficult not because people don’t care, but because when something uncommon and dramatic happens, it carries enormous emotional weight. The urge to act immediately is strong. But our ability to use clear thinking is flawed in systematic and predictable ways. These biases are always present, but urgency and heightened emotion give them more influence over how decisions are made.

A few of these cognitive biases are especially relevant here:

• Availability bias: we tend to overestimate how common or important events are when they are recent, dramatic, or easy to recall. Highly visible incidents can seem far more frequent than objective data supports.

• Outcome bias: we often judge decisions by how they turn out rather than by the quality of the decision-making process. When a bad outcome occurs, the decision behind it is often viewed as wrong, even if it was reasonable at the time and the outcome was a fluke.

• Confirmation bias: we naturally seek out and favour information that supports what we already believe, while discounting information that challenges it. This can make alternative explanations or contradictory evidence easier to dismiss.

• Overconfidence bias: experience helps us recognize how things go wrong, but it does not reliably improve our ability to estimate how often they occur. For rare events, people with experience may feel confident in their risk judgments while still being wrong about the actual level of risk.

  
  

None of these biases make us bad humans. But a problem arises when decisions are made without structures in place to consciously manage them. When biases go unrecognized, people naturally gravitate toward quick and impulsive decisions, based on incomplete information. These biases are exactly why evidence generation and evaluation are trained skills and why they typically involve multiple people with different expertise. Choosing the right study design to answer a specific question depends on formal expertise developed over many years.

Why observational research exists (and why RCTs aren’t always the right tool)

A recurring theme in the current discussion is uncertainty about what kind of evidence should guide safety decisions. Much of that debate reflects a common belief that experimental research using randomized controlled trials (RCTs) represents the “gold standard” for evidence. But that framing only makes sense for certain kinds of questions. When the problem involves rare outcomes, complex and adaptive systems, or ethical and practical constraints that make randomization inappropriate, observational research is not a second-rate option. It’s the most appropriate one.

Agility is a complex, adaptive sport. Dogs, handlers, judges, course designers, and equipment all influence one another. Changing one element, such as obstacle dimensions, doesn’t operate in isolation. It can affect speed, confidence, handling choices, training approaches, course design, and exposure patterns. RCTs are poorly suited to study these kinds of variable conditions. And when there is genuine concern that an “intervention” could increase risk, deliberately randomizing participants to that condition is highly unethical.

In contrast, observational designs are routinely used to study and manage risk in public health, occupational health and safety, sport, transportation, and other high-risk domains. These approaches are built to address real-world complexity. They incorporate exposure data, allow for heterogeneity in participants and contexts, detect unintended consequences over time, and focus on population-level patterns of risk rather than individual outcomes. Rather than tightly controlling conditions, they allow researchers to observe how risk actually unfolds in real-world settings.

In other words, observational research asks whether patterns hold across many dogs, across many handlers, and across different contexts.

Experience vs. evidence: why both matter, but are not interchangeable

Experience is invaluable. It identifies problems, generates hypotheses, and highlights areas where current systems may be vulnerable. People with hands-on experience often notice issues long before formal studies are conducted.

But experience alone cannot answer questions about frequency, comparative risk, or net impact across a population. In other words, seeing a problem occasionally does not tell us how often it occurs, how it compares to other risks, and what it means at a sport-wide level. 

Observational research is the step that comes after experience. It asks: how often does this actually happen; under what conditions; compared to what; and with what trade-offs?

Confusing experience with evidence doesn’t strengthen decision-making. It weakens it. It risks letting those cognitive biases we talked about earlier drive outcomes, leading us to draw conclusions from flawed inputs.

What responsible evidence-building looks like in applied sport safety

Turning experience into evidence requires more than a single study or perspective. Responsible evidence-building is necessarily multi-pronged. No single method is sufficient. What this looks like in practice typically involves several complementary approaches, many of which are already underway.

These include prospective incident surveillance using clear and consistent definitions, so that events are captured systematically rather than anecdotally. Crucially, this surveillance data must be analyzed using exposure-adjusted rates, rather than raw incident counts, because understanding risk depends on how often an obstacle is attempted, not just how often something goes wrong.

In addition, structured video analysis using predefined criteria can help identify common patterns and contributing factors when incidents do occur, while performance and biomechanical characterization can provide insight into how dogs move, load, and adapt under different conditions. Finally, these data need to be interpreted together. No single source tells the whole story, but triangulating across surveillance data, video analysis, and performance measures allows patterns to be evaluated more reliably than any one approach in isolation.

This kind of evidence-building does not produce instant answers. But it does produce more reliable guidance, and reduces the risk that our biases dominate decisions.

Why knee-jerk changes can backfire

In complex systems, well-intentioned changes can have unintended consequences. This is not because people are careless, but because when equipment or rules change, behaviour, biomechanics, and exposure often change with them, sometimes in ways that aren’t obvious at first.

Changes to the dogwalk, such as widening it, lowering it, or removing it, may feel intuitively safer. But intuition is not evidence.

For example, it’s plausible that:

• a wider or lower dogwalk could increase speed and confidence,

• increased speed could increase impact forces when things go wrong,

• handling strategies, training approaches, and course design could shift in ways we haven’t anticipated,

• and risk could move from one failure mode to another, rather than disappear.

  
  

Without data, we don’t know whether the net result is safer, the same, or worse.

We’ve seen this play out before, following both rule changes and equipment changes in other sports. Here are two examples.

Example 1: Youth ice hockey and body checking (rule change)

In North America, rules around body checking in youth ice hockey have been modified repeatedly with the goal of reducing injury risk. One strategy implemented in the early 2000s was to introduce body checking at younger ages, when players are smaller and move more slowly, resulting in lower collision forces. The underlying assumption was that early exposure would allow players to learn how to give and receive body checks in a lower risk environment, leading to a protective effect as they grew bigger, faster, and stronger over time.

What only became apparent after implementation, however, was that this protective effect did not materialize. The data showed that, regardless of the age at which body checking was introduced, injury rates increased significantly and immediately compared with non-contact leagues, and remained elevated year after year. In practical terms, those players introduced to body checking at 9–10 years of age did not experience lower injury risk later. Instead, early exposure resulted in four to five additional years of participation at higher injury risk compared with today’s active players, who are not exposed to body checking until 14–15 years of age.

Example 2: Ski equipment changes in alpine skiing (equipment change)

A parallel example comes from alpine skiing. Over several decades, advances in ski boots and binding release systems led to dramatic reductions in serious lower-leg fractures. This was a major safety success. Unfortunately, at the same time, injury surveillance revealed a rise in serious knee injuries, particularly ACL tears, which became one of the most common severe injuries in the sport. This is a clear case of unintended consequences. Changes to the equipment system anchoring the foot to the ski altered how forces were transmitted through the lower body during falls, reducing injury to the lower leg, but shifting loading to the knee. The injury burden changed rather than disappeared.

As with hockey, this outcome was not obvious at the time the equipment changes were made. It only became clear after systematic data surveillance.

These examples don’t tell us what will happen if the dogwalk is changed. They tell us why it’s risky to assume we already know. Without data, we don’t know whether dogwalk modifications would reduce harm overall, or trade one problem for another.

That’s why decisions driven by urgency, rather than evidence, carry a real risk of creating new problems while trying to solve existing ones.

Consequences if we get this wrong are real

Decisions about agility equipment and rules are not neutral.

They carry financial costs for clubs and organizations, adjustments to training practices for handlers and instructors, (re)learning implications for dogs, and coordination challenges if standards diverge across organizations.

Once implemented, many changes are difficult or impossible to reverse. That should raise the bar for evidence, not lower it. This doesn’t mean change should never happen. It means that change should be justified by clear evidence that the benefits outweigh the foreseeable costs.

Waiting is not inaction

Improving safety in sport rarely comes from a single study, responding to highly visible rare incidents, or relying on one type of evidence alone. It comes from bringing multiple sources of information together and asking what they collectively tell us about risk, trade-offs, and unintended consequences.

Getting this right takes time, not because nothing is happening, but because building reliable evidence requires careful data collection, analysis, and synthesis across methods. Research evidence is not meant to be the sole determinant of decisions, but it plays a critical role in ensuring that debates and policy changes are informed by the best available evidence rather than driven by urgency alone. This is the central idea behind evidence-based practice: integrating research evidence with expertise and context to support better decisions.

Raising concerns about evidence, bias, and unintended consequences is not the same as being opposed to change. Many people engaged in these discussions agree that changes to the dogwalk may ultimately be warranted. The question is not whether change should ever happen, but how it should happen, and on what basis.

Caring about safety and insisting on evidence are not opposing positions. They are the same position, but applied responsibly.