Counterfactuals

If, like me, you grew up during the Friends era, you might remember the scene where Ross is upset because he has to tell Rachel that Emily insists they can’t be friends.

Ross: “I don’t want Rachel to hate me. I don’t know what to do.”
Joey: “You want my advice?”
Ross: “Yes! Please!”
Joey: “You’re not gonna like it.”
Ross: “That’s ok.”
Joey: “You got married too fast.”
Ross: “That’s not advice!”
Joey: “I told you.”

For once, Ross is absolutely right. That isn’t advice. It’s a counterfactual: a redundant statement that sounds wise because the outcome is already known. It’s the illusion of clarity after the fog has lifted, after the complexity has collapsed into certainty. We all know the feeling of wanting to tell someone (or rather often ourselves) what should have been obvious. But that sense of “obviousness” only exists from where we’re standing now, in 20/20 hindsight, not from where the decision was made.

Counterfactuals are seductive. They tidy up messy stories. “If only we’d done X.” “If only they’d noticed Y.” They sound analytical, but they’re fictional. As the saying goes, “If my grandmother had wheels, she’d be a bicycle.” Once we start changing the facts, we’re not talking about reality anymore: we’re imagining a different one that didn’t happen.

It’s easy to laugh at Joey, but we all do it. Just look at all the hot takes on the large AWS outage this week. We look back at a failed project, a near miss, an incident or accident, and feel the seductive pull of “they should have..” or “they shouldn’t have…” because we crave causality and coherence. When something goes wrong, we want to believe that there was a single point of failure that we can fix for next time, reassuring ourselves that it won’t happen again. But as Dekker reminds us, 

“…[counterfactuals] make you spend your time talking about a reality that did not happen (but if it had happened, the mishap would not have happened.)” (Dekker, The Field Guide to Understanding Human Error, 2006, p. 39). 

Counterfactuals don’t actually explain why people acted as they did. 

Hindsight gives us the feeling that we’re seeing more clearly, when in fact we’re seeing differently. It turns messy events into clean, neat, linear stories where the outcome seems inevitable. After the fact, we reconstruct a chain of events so neatly that it feels like it couldn’t have unfolded any other way. But at the time, it could have unfolded many ways.

In complex systems, whether in tech, healthcare, or flying planes, everything seems to make sense until it suddenly doesn’t. Karl Weick describes this as the “collapse of sensemaking”, where a lack of relevant information, conflicting or contradictory data, distractions, or cognitive biases prevent accurate interpretation of the situation. As Richard Cook said, “All practitioner actions are gambles.” Every decision is made under various degrees of uncertainty, using local cues, imperfect data, and competing goals, constraints and pressures. When we uncritically apply hindsight, we risk erasing that uncertainty and turning those gambles into moral judgments. A decision that led to a failure is considered a mistake; but the same decision, if it led to success, is considered high expertise. The same action has a different label depending on its outcome.

This is the core of counterfactual thinking: imagining an alternative world where things turned out differently. Psychologists describe two main types:

• Upward counterfactuals: “We’d have been ok if we had…”, which can be painful but sometimes useful. If we’re careful in their use, they help us see what might be improved for the future.
• Downward counterfactuals: “It could have been worse…”, which might feel comforting and soothing, but is rarely instructive.

Both kinds risk making the world seem more controllable than it really is. Counterfactuals whisper in our ear that different choices would have guaranteed a different outcome. But the real world doesn’t work like that – if we change one variable, the network of interactions shifts in unpredictable, potentially cascading ways. And there’s no control group for the past: we can’t A/B test reality, so we can’t really say that another approach would have turned out differently.

Dekker calls counterfactual reasoning one of the biggest traps in investigation. It’s not that counterfactuals are evil, because used effectively, they can help us think about future prevention – but they can’t explain human behaviour. Saying what people didn’t do tells us nothing about why they did what they did. To learn, we have to reconstruct what the world looked like from their perspective at the time, not ours. To paraphrase Dekker – we need to get in the tunnel with them.

That reconstruction (the “unfolding mindset”) is the work of effective investigation and learning. It means asking what information was available and visible, what explicit and implicit goals were people working towards, what pressures, constraints and distractions were present, and what state the people were in. It’s slower and messier than counterfactual shortcuts, but it’s the only way to understand Local Rationality – why a decision that now looks foolish made sense at the time.

This connects directly to psychological safety. Psychological safety isn’t just about speaking up in the present: it’s about how we talk about the past. Do people feel safe to be candid about their past actions, decisions, perspective and pressures? When we replay an incident, do we talk about what people missed, or do we ask what they saw? Do we focus on what should have happened, or on what actually made sense given their constraints and context? Do we treat decisions as expert gambles rather than tests of character? The way we tell the story shapes whether we learn or blame (and in the majority of cases, we can’t do both).

Put hindsight, counterfactuals, and the fundamental attribution error together, and we get a recipe for unhelpful investigation:

1. Hindsight bias says the outcome was obvious.
2. Attribution error says someone failed to see it because of who they are.
3. Counterfactuals provide the tidy fix: “They should have done X.”

A disturbing number of real incident reports follow this pattern.

The alternative is slower but more honest: reconstruct rather than rewrite. We can attempt to reconstruct the unfolding mindset of the people involved. We can ask what looked risky or safe from their vantage point. As Dekker and David Woods both emphasise, people’s actions almost always make sense in context, even if that context is now invisible to us.

James Reason’s “substitution test” can help here – asking whether another well-motivated, equally competent person might have done the same thing in that situation. 

“Substitute the individual concerned, for someone else coming from the same domain of activity and possessing comparable qualifications and experience. Then ask the question ‘In the light of how events unfolded and were perceived by those involved in real time, is it likely that this new individual would have behaved any differently?’”

If the answer is yes, the issue isn’t the person, it’s the system. However, there’s a big caution here: the substitution test is highly vulnerable to hindsight too because we still have to reconstruct the at-the-time context before we apply the test.

Here are a few practical ways to mitigate counterfactual reasoning and shift our posture from judgment to learning:

• Consider various alternatives to the actual outcome. Explore how it may have unfolded in different ways
• Name the bias early. Simply saying, “Let’s watch for hindsight bias.” can help people notice when it creeps in. It’s not a fix though.
• Reconstruct, don’t rewrite. Ask things like “What did we know then?” and “What options looked viable?” Try to keep current knowledge out until the end.
• Use neutral language. “The decision was made based on X,” instead of “Bob failed to see Y.”
• Broaden the view. Bring in different roles, disciplines, and perspectives. Complexity looks different from other angles.
• Reframe counterfactuals as system questions. Instead of “They should have…” ask “What features of the system made that outcome more likely?”
• Separate accountability from blame. Accountability can coexist with empathy, apologies and amends, but without context, accountability is just punishment.
• Don’t ignore emotional counterfactuals. “I wish I’d done X instead of Y” can be a healthy part of processing regret. It helps us move on before real learning begins.

Joey, like most of us, was satisfied by a simple story, rewritten with 20/20 hindsight. But the real world doesn’t work that way. The goal of learning isn’t to find who “should have known better” – it’s to understand how the system and context made that decision a rational one at the time, and how we can make better outcomes more likely in the future.

Further Reading

• Types of Human Error: https://psychsafety.com/psychological-safety-human-error/ 
• Work as Imagined vs Work as Done: https://psychsafety.com/psychological-safety-work-as-imagined-vs-work-as-done/ 
• Meadows, S., Baker, K. and Butler, J., 2005. The incident decision tree: guidelines for action following patient safety incidents. Advances in patient safety: from research to implementation, 4, pp.387-399.
• Weick, K. (1993). The collapse of sensemaking in organizations: The Mann Gulch disaster. Administrative Science Quarterly, 3: 628–652.
• Woods, D. D., Dekker, S., Cook, R., Johannesen, L., & Sarter, N. (2010). Behind Human Error (2nd ed.). Farnham: Ashgate Pub Co.
• Vaughan, D. (2009). The Challenger Launch Decision: Risky Technology, Culture, and Deviance at NASA. University of Chicago Press.
• Dekker, S. (2003). Failure to adapt or adaptations that fail: contrasting models on procedures and safety. Applied Ergonomics, 34(3), 233—238. doi:10.1016/S0003-6870(03)00031-0
• REASON, J. (1990). Human error. New York: Cambridge University Press.
• Reason J. Managing the risks of organizational accidents. Aldershot: Ashgate Publishing Group; 2003. pp. 191–222.
• Richard Cook. How Complex Systems Fail. https://how.complexsystems.fail/ 
• Why Are All the Smart People So Bad at History? Joan Westenberg. https://web.archive.org/web/20250418033302/https://www.joanwestenberg.com/why-are-all-the-smart-people-so-bad-at-history/ 
• Allspaw, J. Counterfactual Thinking, Rules, and The Knight Capital Accident https://www.kitchensoap.com/2013/10/29/counterfactuals-knight-capital/ 
• Dekker, S. (2006). The Field Guide to Understanding Human Error. Ashgate Publishing, Ltd. https://www.humanfactors.lth.se/fileadmin/lusa/Sidney_Dekker/books/DekkersFieldGuide.pdf 

Psychological Safety in Practice

Stammering and speaking on the phone

For people who stammer, stutter, tic, or have other speech differences, phone calls can be a nightmare. Just the pressure of having to speak at the right time, and the fear that being silent for too long means the call could be dropped, can be overwhelming. Everyone who knows me knows that I absolutely hate speaking on the phone! 

If you stammer, it could mean that when you finally get through to your GP after a long wait, you get hung up on because they think you’re not there. This report shows that many people with a stammer often can’t get through voice recognition systems, and are sometimes hung up on when they stammered on a call.

This amazing campaign by Stamma encourages businesses “Don’t hang up, hang on” and give people time to speak. Find out more here. 

Staying on stammering, this is a powerful episode of the Distraction Pieces Podcast with Scroobius Pip, titled “Dissecting Disfluency” with Cherry Burns). In this, Pip discusses his stammer, and how stammering is not the problem, it’s how the world meets the stammer that is the problem.

Neuroinclusivity and psychological safety

As we so often say, psychological safety looks different for everyone, but most psychological safety models assume that everyone processes feedback, attention, and risk the same way. Nicola Knobel pulled together this great aggregation of neurodiversity and psychological safety research, and put it into this table:

Imagining and implementing a responsibility culture in healthcare

Here’s a great paper by Philip Berry about patient safety and investigations, and trying to find the balance between delivering what many patients wish to see and fostering a culture that is not centred on blame – which seems very hard to achieve.