Context and Cacti
Choosing a meal as a philosophical exercise
The evening before April’s solar eclipse, my then-girlfriend and I took a late-night drive to our favourite vantage point overlooking my hometown. The view is a comforting mix of the forests I grew up exploring, the rim of the lake, and a dotting of evening lights. We chatted, laughed, and enjoyed some much-needed time together after over a month apart.
We decided to extend the evening with some food. She had the great idea of getting nachos at a nearby Tex-Mex restaurant called Thirsty Cactus. Splendid! It was exactly the greasy midnight snack that the moment called for.
We looked it up to check the hours and quickly discovered they had already stopped serving food for the night. Not-so-splendid. We quickly looked up alternatives, settled on some other snack, and called it a night.
The next day, we were deciding where to grab a special eclipse lunch before driving to our viewing spot an hour away when my girlfriend excitedly exclaimed — “Thirsty Cactus!”
In this endearing and unexpectedly generative moment, she had demonstrated the fallacy I now call the Cactus Effect, or cactusing for short.
Formally, cactusing is when you make a correct decision in one context, then the context changes in such a way that should impact your decision but you fail to update your decision. The purest form of cactusing is when the new context alters the outcome goal of the decision as opposed to the facts of the situation; in this way, it is distinct from a lack of updated Bayesian priors. To further emphasize this difference, cactusing is when your decision isn’t updated — a decision being a subjective judgment based on specific motivations — as opposed to a descriptive claim about the world modelled through distributions. It’s not that the model’s variables need tweaking but an altogether different measurement is required.
With my girlfriend's original restaurant decision, the original outcome goal of “choosing a restaurant” was to find a late-night indulgence, the kind of food you can share easily and linger on deep into the night. To this end, she correctly chose nachos. The next day, the goal had changed to a quick meal on the road that had eclipse symbolism — but she didn’t update her decision accordingly.
Instead of getting nachos, we discussed the new context and decided on hamburgers. Not only was the meal quick and delicious, it was the perfect eclipse meal, complete with satiating symbolism: the burger begins as a perfect circle, representing the complete sun. Each bite enlarges a growing negative space, representing the crescent of lunar encroachment. Eventually, nothing remains…totality!
“In the stillness and the darkness, realization soon began to supplement knowledge. The mere knowledge of a fact is pale; but when you come to realize your fact, it takes on color.”
— Mark Twain on solar eclipses in A Connecticut Yankee in King Arthur's Court
After that day, I began to see examples of cactusing everywhere: a friend stubbornly returning to a once-great restaurant that had declined under new ownership, a neighbour still treating his son like a little boy even though he had become a grown man.
My favourite was when my friend posted a song he was trying to promote to an internet forum that used to be a very popular place to share music, but has since become completely inactive. What makes this so delightful is that he had washed dishes at the actual Thirsty Cactus restaurant in high school!
I also started to notice trends and clusters of cactusing. One such cluster is status cactusing, or the irrational estimation of social status using past contexts rather than present-day reality. You can see this in both individuals (“the guy who peaked in high school” trying to relive the glory days) and organizations.
An egregious example of organizational status cactusing is with NASA. It was once the coolest place around: the absolute frontier of engineering, commanding 4.4% of the entire federal budget, taking giant leaps for mankind. Today? NASA hardly even takes small steps. Forget about being the top engineering organization in the world — it isn’t even the best at space engineering. Reduced to a measly 0.5% of the federal budget and stripped of any grand ambitions, bureaucracy has supplanted greatness. These contexts shifts ought to have changed NASA’s outcome goals from “world-leading engineering organization” to “marginal American scientific administration”, along with a corresponding difference in treatment.
However, some people still hold NASA in inappropriately high regard because they cling to its past greatness. Current employees take advantage of this fact and brag about their jobs there as if they personally stepped foot on the moon in 1969. They are trying to elicit a social cactus response in you so they can rest on stolen laurels.
You might argue that evaluating people and organizations based on their track record is rational, and you’d be right — if the conditions under which their track record was created persist today! None of the scientists, engineers, or astronauts who made NASA great are still there today, to say nothing of the processes and programs they established. Are there still some talented individuals at NASA? Almost definitely! But they need to be evaluated in their own right instead of inheriting the assessment of their forefathers.
The Dynamics of Cactusing
Though cactusing is when you stick to decisions made in old contexts irrationally, there are sometimes rational justifications to do so. However, these justifications can be taken too far and lead you to fallacy once more. Think of cactusing as a disease, the justifications as the drugs that treat the disease, and the return to fallacy as an overdose on those drugs. Here are the possible justifications:
An emotional attachment, either to the original context (nostalgia), the original decision (stubbornness), or the outcomes of the original decision (chasing the dragon). This may be justified if the value of the emotional attachment outweighs the marginal benefit of updating the decision, but can lead to overdose if you let your emotions rule in moments where rationality must prevail. Charitably, attachment to the original context is what my girlfriend’s suggestion of going to Thirsty Cactus was — holding on to the tender moment we shared the previous night.
The cost of updating your decision outweighs the marginal benefit of the update, creating a natural decision inertia. This is evidently justified, but can overdose into the sunk cost fallacy if you overestimate update costs.
It could be that the original decision was made by somebody (or a past you) who had more complete knowledge about the original context than you do now, and by maintaining the original decision you are exercising cautionary principles à la Chesterton’s Fence. This is important in decisions with large downside risk and small upside potential, going doubly if the decision is irreversible (i.e. there is a light switch with a label “DO NOT TOUCH!” ). It can be overdosed if you don’t account for a big risk of not updating your decision, let alone reversible decisions with low risk and high upside where such principles don’t apply at all (i.e. trying a new haircut). A further overdose may arise if you overestimate the difference between the original decision-maker’s insight into context and your current insight.
Realistically, most perceived instances of cactusing are the original irrational thing, caused by:
A lack of awareness, either that the context has changed at all, or about the way the context has changed that is material to your decision.
(Uncharitably, this is what my girlfriend’s suggestion of going to Thirsty Cactus was — not realizing how our needs had changed.)
These dynamics of cactusing, with all these different justifications and higher-order miscalculations, can be quite chaotic. And in this beautiful chaos lies opportunity:
The most important takeaway of cactusing is its role in generating usefully exploitable private information in Thiel’s “secret knowledge” sense as described in his book Zero to One. The dynamics of cactusing offer new mechanisms to generate the unique insights that he describes as the main source of an advantage, or what hedge fund managers might call their alpha.
Without the Cactus effect, the only way to generate secret knowledge would be to be the best at making a decision in a given context, loosely describable as “being the smartest in the world.” In this scenario, everyone would be on level ground, accelerating on a straight track, competing in a race where the fastest car wins.
But given cactusing, there exist decisions in the world that have been made in contexts that are no longer relevant. If you find these decisions and identify the relevance of the new context, you can create secret knowledge in a whole new way — even if you aren’t the “smartest in the world.” The Cactus effect introduces turns in the racetrack, ways in which you can still pull ahead and win the race even if your car isn’t the fastest on a straight track. Some examples of companies that capitalized on a specific moments of context change are Uber with smartphones, Amazon with the Internet, and Salesforce with cloud, all of whom ran circles around incumbents who were still operating based on old paradigms.
Interestingly, these dynamics can create avenues for alpha even if the incumbent is aware the context has shifted! Recall that there are rational justifications for not updating your decision even if you’re aware of a new context, most importantly the cost of updating. This creates a natural cycle where incumbents choose not to adapt to evolving contexts, creating opportunity for new players. Thus, knowing when this window of opportunity opens is yet another way to generate alpha even if you aren’t the “smartest in the world” and even if you don’t even have completely unique insight about a context shift! An example of this is how, despite a paradigm shift towards AI chatbots, Google is unable to abandon classic Search due to asymmetric opportunity costs (it would cost Google $175B/year to abandon Search but costs OpenAI $0 since they never had a search engine in the first place; more on this Incumbent Paradox in the comments).
These opportunities are most relevant in complex domains with many higher-order considerations. The more complex and high-order the consequences of a context shift, the less likely it is for decisions to be updated correctly, so the comparative benefit of recognizing the proper relevance of the new context is higher.
(I have much more to say more about the turns in the racetrack and how complexity creates opportunity in specific, exploitable patterns. Stay tuned for future essays and a longer-form project.)
With all this said, now is as good a time as any to reconsider some old choices you may have made in no-longer-valid contexts — might you be cactusing on some important decisions?
A question best considered over a plate of nachos.
There’s a related idea that isn’t quite worth a full essay:
I want to expand on the point about asymmetric opportunity cost as in the Google example to illustrate what I call the Incumbent Paradox.
Imagine an incumbent company in a tumultuous industry that is faced with a potentially-market disrupting paradigm shift every 5 years. The shift has a 75% chance of being ultimately irrelevant and a 25% chance of changing things enough to make the incumbent’s product obsolete.
Assuming the company can only pursue one paradigm, the rational move is to ignore every individual paradigm shift. It obviously makes sense to take the 75% chance of status quo dominance instead of risking it all on a pivot for a 25% chance to ultimately only reach the same market position they have now. The paradox is that despite only making rational decisions, over 30 years, there is only an 18% chance they won’t have been made obsolete [P(X = 0) for X ~ Bin(6, 0.25)].
This means that, given these assumptions, even fully rational and competent incumbents will eventually fail because of asymmetrical opportunity costs in moments of flux. The average time-to-failure can be approximated using the time between paradigm shifts (t) and the probability of obsolescence from each shift (p) in the simply relationship t/p. For a slow-changing company like the Guinness brewery, you can be around for hundreds of years. For a tech company like BlackBerry, not so much. In practice, companies can usually challenge the formulation and stay alive by hedging on new paradigms through acquisitions and internal R&D (Google AI/X, MSR, FAIR, etc.) or by having more accurate estimates on the likelihood of disruption. But not always! Sometimes, you’re forced to choose between two paths forward.