🔮 Uncertainty as Doubt (Part 1)

Good morning!

At the Uncertainty Project, we explore models and techniques for managing uncertainty, decision making, and strategy. Every week we package up our learnings and share them with the 2,000+ leaders like you that read this newsletter!

This week and next we build on the learnings from John Kay and Mervyn King’s “Radical Uncertainty” with a review of ideas from Amar Bhide’s “Uncertainty and Enterprise”.

It is surprisingly difficult to craft a definition of uncertainty. 

Because of that, it is illuminating to study how different definitions were formed, changed, and debated over time.

This week and next, we’ll take a tour of how uncertainty, as a concept in business, over the last 100 years. Our tour guide will be professor Amar Bhide, who has taught at leading business schools over his long career. His recent book “Uncertainty and Enterprise” does a great job of resurfacing some landmark definitions of uncertainty, describing their evolution through the 20th century, and launching his own effort to pragmatically “modernize” these definitions, for today’s needs.

His tour begins with a re-introduction of Frank Knight. His article “Uncertainty, Risk, and Profit” (1921) was one of the first to draw a contrast between risk and uncertainty:

• Risk can be objectively calculated, from frequencies, historical statistics, or probabilities like a coin toss

• Uncertainty is defined by exclusion: things that we cannot measure in these ways

Bhide is interested in how definitions of uncertainty shape the risk-taking in enterprise settings. That was Knight’s interest as well. 

They both ask the fundamental question: 

 For an entrepreneur, “Where do profits come from?”

These judgments about “what will happen” are made in the presence of uncertainty (per Knight’s definition), not risk. So the use of probabilities, in Knight’s view, was not appropriate.

Years later, Jimmy Savage (now that’s an all-time great name!) established “subjective probabilities”, by which an individual can attempt to state a probability about uncertain things “as if” they knew the frequencies or historical statistics. This idea seems a little crazy, but it opened the door to a more mathematical approach to modeling uncertainty, in economic circles. This reliance on subjective probabilities became the norm for mainstream economic thinking (and spawned tools like decision trees).

This shift towards the use of subjective probabilities (like with Bayesian Thinking) pushed Knight into the background of mainstream economic theory. 

Bhide thinks he deserves a second look.

Bhide describes his efforts in the book as a “modernization project” for Knight’s work - driving a shift from treating uncertainty as the absence of historical frequencies (per Knight) to treating uncertainty as the mental state of doubt. 

The new idea here is that uncertainty could be approached as a personal (“subjective”) mental state. He argues that uncertainty is “psychological, located within the individual” and that Knight’s definition as “situational uncertainty” makes it impossibly objective - and fails to acknowledge that it’s the individual that decides whether something is uncertain.

By shifting the focus to doubts, we change the conversation from “what is” (objective) to “what we feel it is” (subjective), but avoid the crazy reliance on probability. Doubts can emerge from our individual (or collective) ignorance of existing or past conditions, i.e. the “known unknowns”. Having doubt is the opposite of having conviction, which is how we (generally) feel in the absence of uncertainty. 

Bhide describes how uncertainty is handled in legal circles, to draw a contrast with how it is handled in economics. In economics, general, universal theories are held in the highest regard. In legal settings, it’s more pragmatic, and this offers a clue to better definitions of uncertainty.

He reminds us that “beyond reasonable doubt” prosecutions do not remove all possible doubt. But… with doubt comes the possibility of error, which is key. As Bhide says, “The possibility of error raises issues of justification, which are at the heart of my project to update Knight.” 

Here’s an illustration of how evidence (or a lack thereof) can drive us across a spectrum from confidence to doubt. It shows two styles of evidence, echoing Knight’s distinctions between contexts predominately facing risk (shown on the bottom) and uncertainty (shown on the top).

When historical statistics are available (revealing expected frequencies), we talk about how sample sizes might influence our position on the spectrum from confidence to doubt.

When historical statistics are not available (true uncertainty, per Knight), we talk about how pieces of contextual evidence help us build plausible stories about the unknowable future. The volume, quality, and shared understanding around that evidence influences where we position ourselves on the spectrum from confidence to doubt.

So what does it take to reduce our doubts, in the name of reducing uncertainty? Our tour will look at this from several angles:

• Agreement - involving other people

• Justification - making a case

• Evidence - volume, quality, and interpretability

• Ambiguity - degree of interpretability

• Large Organizations - and their routines (covered next week)

• Narratives - two modes, and good form (covered next week)

In general, we seek to reduce our doubts by gaining agreement with others, using narratives that provide justifications based on evidence - that is often steeped in ambiguity. In large organizations, this is often supported by specific routines.

On Agreement

Bhide frames his definition of uncertainty in the “known unknowns”, avoiding the existential angst around “unknown unknowns” (saying those conversations are usually a waste of time).

For the “known unknowns”, we are forced to grapple with our doubts about “what might be”. And this grappling is best done with others. He says, “Where there is doubt, there is the possibility of disagreements.”

Sometimes these disagreements can just be traced back to conflict of interests. But there are also cases in discussions of a “one-off” decision (e.g. a unique, unprecedented business opportunity), where interests are aligned, and disagreements still emerge.

So when should we expect disagreement?

• When information is incomplete

• When situations are new or novel

• When attitudes and temperament are different

Not every disagreement is worth a fight. But when the stakes matter, it “can spur efforts and arrangements to reach agreements through justificatory discourse - the giving and taking of plausible reasons.” And - most importantly - it is “the magnitude of the stakes - the cost of a mistake or misjudgment - (that) affects the conventional demand for justification.”

On Justification

But where do we find the supply for this demand for justification? 

We supply it via our routines (i.e. rituals, meetings, messaging threads, hallway conversations).

High stakes decisions drive routines that “embody a high demand for verifiable evidence.”

Crucially, these routines “often require extended deliberation, not just more evidence.”

When collecting evidence, we face the law of diminishing returns: “This question of sufficiency - deciding how much evidence is enough and when to stop looking for more - also poses acute challenges.”

Bhide, like Olivier Sibony and the concept of decision architecture, points to organizational routines as the way to address this challenge. On this topic, he does not find much to recommend in mainstream economic theory:

“My answer unapologetically focuses on ‘pragmatic’ domain-specific routines that economists and decision theorists often ignore. Here, as in many practical matters, we must manage the problems of relevance and sufficiency without much guidance from a general theory.”

Instead, he recommends following the pragmatism of Herbert Simon, and his idea of “satisficing”.

On Evidence

A few weeks ago we shared John Kay and Mervyn King’s take on the impossibility of building complete decision trees.

Herbert Simon (who we discussed here last year) felt the same way above this chasm between the theory and our actual behaviors. 

In general, Simon “regarded businesses and other organizations as ‘machinery for coping with the limits of man’s abilities to comprehend and compute in the face of complexity and uncertainty.’”. Now, that’s a solid definition!

Bhide describes what led to Simon’s concept of “satisficing”, which is his alternative to optimizing around utility- or profit-maximization:

“Standard global rationality required calculating all possible payoffs. But there was no evidence that ‘in actual human choice situations of any complexity, these computations can be, or are in fact, performed.’ Actual decision making, therefore, required simplification.“

Simon proposed that the simplification occurred thus: ‘If the alternatives for choice are not given initially to the decision maker, then [s]he must search for them.’ And as with evidentiary weight accumulation… (this) searching requires rules for stopping. But someone who starts searching because she does not know what alternatives are available cannot know when she has found the best possible one. 

Simon proposed a simple rule: Evaluate alternatives against an aspiration level, not an unknowable maximizing standard. When an alternative ‘satisfies’ the aspiration, stop; otherwise, continue searching.”

[As a side note: This emphasis on setting aspirations to guide an exploration of a possibility space is similar to how Roger Martin starts his “Strategy Cascade” with a definition of a “winning aspiration.” ]

Simon also built organizational theories that envisioned organizational procedures for dealing with the “social-psychological factors” could impede decision making, like:

• Replacing global profit-maximizing goals with satisficing subgoals

• Dividing the decision-making task among many specialists

• Coordinating the work of these specialists by means of a structure of communications and authority relations

In Simon’s view, an organization should strive to decentralize the decision making, and aim for solutions (and levels of evidence) that are “satisficing”, not optimized:

“Standard procedures - in legal trials, diagnostic medical protocols, reviews of mortgage applications, and employment reference checks - require efforts to secure satisfactorily complete information. Although they vary in the extent of their uncertainty aversion (and never expect totally complete information), the routines are certainly not uncertainty-neutral.” 

This point - that most economic models claim to be “uncertainty-neutral” - is what drives Bhide up the wall. He is less agitated by the challenges of building a complete decision tree (unlike Kay and King) but more frustrated that these economic theories continue to pretend that uncertainty does not exist.

“How much omniscience does utility-maximizing behavior require? Plausibly, mainstream theory merely demands ‘practical omniscience’... Theoretically rational decision-makers don’t have to be oblivious to the incompleteness of their information. They can maximize based on what they know, cooly treating what they do not as ‘neutral’ noise.”

So while these mainstream economic theories claim to be “uncertainty-neutral”, Bhide is pushing for being more than just “neutral”. 

On Ambiguity

Next (as if defining uncertainty wasn’t hard enough), we find ourselves having to talk about the uncertainty-surrounding-the-evidence-we-collect-to-reduce-the-uncertainty!

This is where the concept of ambiguity has been introduced. Bhide summarizes the leading thinkers on this topic:

• Daniel Ellsberg defined ambiguity as the “quality depending on the amount, type, reliability, and ‘unanimity’ of information, giving rise to one’s degree of ‘confidence’ in an estimate of relative likelihoods.”

• Colin Camerer says it surfaces when we are “known-to-be-missing information, or not knowing relevant information, that could be known.”

Bhide notes that, “Most decision theorists today consider ambiguity as a synonym for uncertainty. A minority, however, maintain that ‘fundamental uncertainty pertains to situations in which information does not exist at the time of a decision, while ambiguity refers to missing information that could be known.”

And known-to-be-missing information has psychological ramifications. Ellsberg said, “Not knowing important information” can be “upsetting and scary” making “people shy away from taking either side of a bet.” This fear is what leads them to drag their feet when trying to establish their subjective probabilities.

The psychological ramifications extend to innate qualities from person to person. Consider the stereotypical differences between an entrepreneur and a banker:

“Differences in attitudes towards missing information can hinder joint activity even when there are no differences in information or interests. The same missing information that excites an entrepreneur may repulse potential financiers. What’s more, aligning financial incentives cannot close this psychological/behavioral gap: it requires reducing the amount of missing information, either naturally with time, or by securing more information.”

The key point here is that collecting evidence (usually) won’t be enough to remove doubt, in a business setting. Since doubt is personal, we will bring all of our idiosyncratic selves (and our assumptions about the world) into a group conversation about this unknowable future (and the one-off decisions that chart our path into it):

People are messy. And more people, more mess.

Next week (in Part 2), we will finish this conversation by considering how the context changes when people are part of larger organizations, and exploring how organizational routines, when they lean into narratives, can cut through the mess.

Have a great week!