Leadership and organizational behaviour · Module 1

Leadership under uncertainty and executive decision-making

How should executives decide when the odds cannot be known, the stakes are high and their own judgement is part of the problem?

Beyond the IB course UndergraduateMBAProfessional About 18 minutes · 2,999 words · checked 16 November 2026

From the IB course to this module

The IB course treats leadership mainly as a choice of style (2.3) and crisis as something to plan for (5.7). This module starts where those leave off, with decisions whose probabilities cannot be known. It separates risk from uncertainty, shows how experienced judgement fails in predictable ways, and sets out the processes top teams use to decide well without knowing, at a depth well beyond the course.

Section 01: Conceptual foundation

Every executive decision is a bet on an unknown future. What sets leadership decisions apart is the kind of uncertainty involved. In 1921 the economist Frank Knight drew the line that still organizes the field. Under risk, outcomes and their probabilities can be known from data or from the structure of the situation: an insurer pricing car policies, a lender estimating defaults. Under uncertainty, they cannot: nobody can defend a probability that a new battery chemistry will be cost-competitive in 2032, or that a government will keep a subsidy for a decade. Under ambiguity, even the options are unclear.

The tools differ. Under risk, analysis can find the option with the highest expected value. Under uncertainty, probabilities are judgements, so a precise expected value can hide a guess, and decision quality depends more on the process: how options were generated, how evidence was challenged, how reversible the commitment is and how fast the organization will notice it is wrong.

Two further ideas run through the module. A good decision can have a bad outcome, so judging leaders by results alone teaches the wrong lessons. And leading under uncertainty is more than choosing: it means making sense of an unclear situation for others and keeping them committed while staying ready to change course.

Section 02: Advanced explanation

Risk, uncertainty and the value of keeping options open

Decision theory answers Knight in two ways. The subjectivist view says a rational person can always attach personal probabilities, so uncertainty is risk with softer numbers. The Knightian view says pretending to know odds you do not know is itself an error. Ellsberg (1961) showed that people prefer bets with known odds to bets with unknown odds even when the subjectivist calculus says they should not care: ambiguity aversion is a stable feature of choice.

Under deep uncertainty three principles weigh more than optimization: prefer robust choices that do acceptably in many futures; value reversibility, because an irreversible commitment destroys the option to wait and learn, an option Dixit and Pindyck (1994) showed has measurable value; and buy learning early with cheap experiments.

A worked example. A firm can build a plant now for £100m, earning £220m in present value if demand is high and £60m if it is low. Or it can spend £20m on a pilot, then build for £90m if demand proves high, earning £200m because it arrives a year later, or stop and recover £10m if demand is low. With p the probability of high demand:

  • Commit now: expected value = 220p + 60(1 − p) − 100 = 160p − 40.
  • Pilot first: expected value = (200 − 90)p + 10(1 − p) − 20 = 100p − 10.

The two break even at p = 0.5, so commitment pays only if high demand is more likely than not. The extremes matter as much: if demand is low, commitment loses £40m and the pilot £10m. When p is a guess, that asymmetry often decides.

How expert judgement goes wrong, and when it does not

Tversky and Kahneman (1974) showed that people estimate uncertain quantities with shortcuts that are usually useful and sometimes badly wrong. Anchoring pulls estimates towards any number in view; availability makes vivid events seem more likely; representativeness leads people to judge by resemblance and neglect base rates. Executives add overconfidence: their 90% confidence ranges contain the truth far less often than 90% of the time.

Two errors matter most for strategy. The planning fallacy is forecasting from the details of one's own project (the inside view) rather than from the record of similar projects (the outside view); Kahneman and Lovallo (1993) argued that managers pair bold forecasts with timid choices. Reference class forecasting, starting from the outcomes of comparable projects, is the standard correction. Loss aversion, from prospect theory (Kahneman and Tversky, 1979), holds that outcomes are judged as gains and losses from a reference point, that losses weigh more, and that people gamble to avoid a sure loss. Applied to a failing project it predicts escalation of commitment, which Staw (1976) demonstrated: people who made the original decision invested more after bad news than people who had not.

The biases school has a serious rival. Researchers in naturalistic decision making, led by Gary Klein, found fire commanders deciding well and fast by recognizing patterns. Kahneman and Klein (2009) reconciled the views: trust intuition only where the environment is regular enough to learn and feedback is quick and clear. An anaesthetist meets both conditions; a chief executive judging a once-a-decade acquisition meets neither.

Deciding as a top team

Most consequential decisions are made by groups, which add failures of their own. Janis (1972) described groupthink: a cohesive team under a directive leader suppresses doubt and stops looking for alternatives. The idea rests mainly on historical cases and later tests have given it mixed support, but the practical lesson stands: teams pool information badly unless the process forces dissent into the open.

Eisenhardt (1989), studying firms in fast-moving computer markets, found that the fastest top teams used more real-time information than slow ones, weighed several alternatives at once and sought agreement, but let the leader decide if none came. Speed came from parallel options, not from skipping analysis.

Several techniques target known faults:

  • The pre-mortem (Klein, 2007): the team imagines the decision has failed and each member writes down why, which makes doubt legitimate.
  • Devil's advocacy and dialectical inquiry assign people to attack the plan or build the strongest alternative.
  • Independent estimates before discussion, so the first number spoken does not anchor the rest. This also cuts noise, the scatter between judgements that Kahneman, Sibony and Sunstein (2021) argue costs as much as bias.
  • Review questions (Kahneman, Lovallo and Sibony, 2011): has the team fallen in love with its proposal, were real alternatives considered, is the base case anchored on an old plan?

Leading people through uncertainty

In uncertain situations people look to leaders to explain what is happening, and the explanation shapes what they do. Weick's (1993) study of the Mann Gulch fire, where most of a smokejumping crew died after their understanding of the situation collapsed, made sensemaking central to organization studies: when the frame people use to read events fails, coordination fails with it.

Heifetz (1994) separated technical problems, which experts can solve with known methods, from adaptive challenges, which require people to change habits or loyalties. Announcing a solution to a market shift that in fact demands new ways of working treats an adaptive challenge as technical, a common failure. Adaptive leadership keeps enough pressure on for people to move, but not so much that they freeze, and gives the work back to those who must change.

Commitment cuts both ways. Ghemawat (1991) showed that costly, hard-to-reverse commitments are what make a strategy credible; under uncertainty the same irreversibility is dangerous. Leaders manage the tension by separating commitments meant to signal from bets on a forecast, by naming signposts that would trigger a review, as the scenario planner does, and by agreeing kill criteria before anyone has a sunk cost to defend.

Honest ranges earn more lasting trust than false precision. Weick and Sutcliffe (2001) found that high-reliability organizations share habits that transfer to business: preoccupation with small failures, reluctance to simplify, and deference to expertise rather than rank in a crisis.

Section 03: Models and frameworks

Where each comes from, how to use it, and where it stops being useful.

Knight (1921); Dixit and Pindyck (1994)

Risk, uncertainty and staged commitment

Sort the unknowns in a decision into risks, with defensible probabilities, and uncertainties, where probabilities are judgement. Where uncertainty dominates and the investment cannot be reversed, the option to wait, stage or pilot has value, and that value rises with the uncertainty.

How to use it

Name the two or three unknowns that most move the outcome. For each, ask what could be learned before committing, how long it would take and what it would cost. Compare committing now with a staged path, find the probability at which they break even, and stage if honest estimates straddle it.

Where it fails

Waiting is not free: rivals may pre-empt the market and partners may go elsewhere, and options thinking can become an alibi for never deciding. The formal method assumes the option will be exercised rationally, which organizations under political pressure often fail to do.

Frank H. Knight (1921). Risk, Uncertainty and Profit. Houghton Mifflin. Avinash K. Dixit and Robert S. Pindyck (1994). Investment under Uncertainty. Princeton University Press.

Tversky and Kahneman (1974); Kahneman and Tversky (1979)

Heuristics, biases and prospect theory

Judgement under uncertainty relies on shortcuts, anchoring, availability and representativeness, that produce predictable errors. Choices are framed as gains and losses from a reference point; losses weigh more than equal gains, and people take risks to avoid a sure loss.

How to use it

Use it as a review checklist for any large proposal. Where did the base-case numbers come from? Do vivid recent events drive the risk assessment? Does the forecast start from comparable projects? Is the project continuing because of what has already been spent?

Where it fails

Effects measured in the laboratory vary in size in the field, and knowing about a bias does little to remove it. The theory explains errors better than it prescribes choices, and a checklist is only as good as the honesty of the people completing it.

Amos Tversky and Daniel Kahneman (1974). Judgment under Uncertainty: Heuristics and Biases. Science, vol. 185, no. 4157, pp. 1124 to 1131. Daniel Kahneman and Amos Tversky (1979). Prospect theory: an analysis of decision under risk. Econometrica, 47(2). Daniel Kahneman, Dan Lovallo and Olivier Sibony (2011). Before you make that big decision. Harvard Business Review, 89(6).

Vroom and Yetton (1973)

The participation model

Five decision styles run from deciding alone, through consulting individuals or the group, to letting the group decide. The right style depends on how much quality matters, where the information sits, whether others' acceptance is needed and likely, and whether they share the organization's goals.

How to use it

Before a decision, answer the diagnostic questions: Is there a quality requirement? Do I have the information? Is acceptance critical, and would a decision I make alone be accepted? Do people share the goals? The answers rule styles out; time pressure usually settles the rest.

Where it fails

Support comes mainly from managers' recall of their own decisions, which flatters the model. It treats decisions as single events, though strategic choices unfold over months, and time costs were added only later (Vroom and Jago, 1988).

Victor H. Vroom and Philip W. Yetton (1973). Leadership and Decision-Making. University of Pittsburgh Press. Victor H. Vroom and Arthur G. Jago (1988). The New Leadership: Managing Participation in Organizations. Prentice Hall.

Section 04: Real business applications

Real organizations appear only with dated, cited facts; illustrative examples are labelled.

Real organization, 2025-02

bp: from transition targets to a reset (2020 to 2025)

In August 2020 bp set out a strategy to become an integrated energy company, with aims for 2030: low carbon investment up about tenfold to around $5 billion a year, around 50GW of net renewable generating capacity developed, and oil and gas production expected to fall by at least one million barrels of oil equivalent a day, or 40%, from 2019 levels.

On 26 February 2025 bp announced what it called a fundamentally reset strategy: oil and gas investment up to about $10 billion a year, production growing to 2.3 to 2.5 million barrels of oil equivalent a day in 2030, and transition investment down to $1.5 to 2 billion a year, more than $5 billion below previous guidance. All aims and targets first announced on 4 August 2020 were retired; the chair cited significant changes in energy markets.

Interpretation. The central unknown in 2020, the pace of the energy transition, was a Knightian uncertainty on which informed forecasters disagreed. Precise public targets made the strategy credible but turned a forecast into a commitment that was costly to reverse. The case does not show which strategy was right; it shows the price of attaching firm promises to a contested forecast without signposts for review.

Real organization, 2025-03

Northvolt: several gigafactories before the first one worked (2024 to 2025)

In September 2024 Northvolt, the Swedish battery maker, announced the first outcomes of a strategic review: it would focus on cell manufacturing at its Northvolt Ett gigafactory, mothball a cathode materials plant and end its Northvolt Fem programme, while staying committed to projects in Sweden, Germany and Canada. In November 2024 it filed for Chapter 11 reorganization in the United States, giving access to about $245 million of new financing, and its co-founder stepped aside as chief executive. On 12 March 2025 its board filed for bankruptcy in Sweden, citing rising capital costs, geopolitical instability, supply chain disruption and shifting demand, and significant internal challenges in ramping up production, some of them unforeseen.

Interpretation. The firm held several large, irreversible commitments in parallel while the yield of its first plant was unproven, the opposite of the staged path in the worked example. Two cautions: in a race for customers and subsidies waiting also carried risk, and this account rests on the company's own statements, so it is a lens, not a finding about the firm's decisions.

Illustrative

A pre-mortem before a hospital system goes live

A hospital is two weeks from moving its patient records to a new system. The programme director tells the team that the go-live has failed and patients have been harmed, and asks each member to write down why. The answers cluster around three causes: night staff were trained least, a pharmacy interface was tested only with clean data, and nobody had agreed when to revert to paper. The board adds a ward-by-ward go-live, a test with real prescriptions and a kill criterion agreed in advance. None of this needed new analysis; it needed permission to voice doubt.

Section 05: Case studies

Fictional case, simulated figures

Thornbury Buses: electrify now, pilot, or wait?

Thornbury Buses, a fictional regional operator, must decide what to do with 120 diesel buses that reach the end of their planned life within three years. Option A replaces the fleet with battery-electric buses now. Option B extends the diesels' life and runs a two-route pilot of battery and hydrogen buses, deciding in three years. Option C extends the diesels and revisits the question in three years without a pilot.

The result turns on public policy, which nobody at Thornbury can forecast with confidence. The finance team has valued each option in three policy scenarios, and three directors have each written down their own probabilities. The chief executive will open the board discussion.

Net present value by option and scenario (£m, ten years), with each director's probabilities
Row Fast policy push Steady path Policy stall
A: electrify now (£m)4015−20
B: extend and pilot (£m)1085
C: extend and wait (£m)−25012
Chief executive's probabilities0.50.30.2
Finance director's probabilities0.20.40.4
Operations director's probabilities0.30.40.3

Simulated figures. Values are estimated net present values at the company's cost of capital; probabilities are each director's own judgement.

Questions

  1. Calculate the expected net present value of each option under each director's probabilities. Which option would each director choose?
  2. Find each option's worst outcome and its maximum regret, the shortfall against the best option in each scenario. Which option is most robust, and what does that robustness cost in expected value on the operations director's view?
  3. On the operations director's probabilities, what is the most Thornbury should pay to know in advance which scenario will occur? What information could it actually buy, and why is that worth less?
  4. The chief executive plans to speak first. Which errors does that invite, and how would you run the meeting instead?

Section 06: Critical evaluation

What the evidence supports, where the debate is, and what the models leave out.

What is established. The main heuristics and biases replicate well: anchoring was among the most reliable effects in a large multi-laboratory replication project (Klein and others, 2014), and loss aversion appears in many settings, though its size varies with context. The planning fallacy is well documented for large projects, and the outside view improves forecasts.

Where caution is needed. Psychology's replication problems apply. A coordinated attempt to repeat 100 published studies (Open Science Collaboration, 2015) obtained significant results in well under half, so a single striking study, especially a small one, is a hypothesis rather than a finding. Groupthink rests largely on Janis's reading of historical cases. Popular practitioner frameworks, such as Snowden and Boone's (2007) Cynefin, which sorts situations into clear, complicated, complex and chaotic, are useful vocabularies rather than tested theories.

What is contested. The two schools still disagree about how far to trust experienced judgement; Kahneman and Klein's conditions reconcile them but leave executives in an awkward spot, since the most strategic decisions are the least learnable. Scholars also dispute whether ambiguity aversion is an error or a sensible hedge against a wrong model.

What the models leave out. Decision models picture a person choosing. In organizations decisions also emerge from politics, routines and incentives, and some are never explicitly taken. And the evidence about leaders is bent by outcome bias and survivorship: bold bets that worked are remembered as vision, identical bets that failed as bad luck. Of any celebrated decision, including those above, ask what was known at the time and how many similar bets failed.

Section 07: Decision exercise

A fictional situation. Decide before you read the guidance.

Publish a date, or publish a path?

Fernlea Distribution, a fictional pharmaceutical wholesaler with 900 delivery vans, is preparing its first transition plan. Its largest customers want suppliers with credible targets, and a van maker has offered a discount for a firm order. But van prices, grid capacity at three depots and government grants are all uncertain. The board must decide what to commit to publicly at next month's investor day.

  1. A. A firm date

    Announce an all-electric fleet by 2030 and place a large order now to secure the discount. Most credible and cheapest per van, but hardest to reverse.

  2. B. A staged path

    Announce the 2030 ambition with published milestones and the conditions that would trigger a review (grid connections, van cost, grants). Order for the first two depots now.

  3. C. No public target

    Electrify depot by depot as the economics allow, with no public commitment. Fully flexible, but customers and staff see no plan and the discount is lost.

Judge the options on

  • Credibility with customers, staff and partners
  • Flexibility if grid, cost or policy assumptions change
  • Reputational cost of missing or retiring a target
  • Financial value, including the order discount
Read the guidance

Ask first what the commitment is for. If its value lies in coordination, persuading customers to sign longer contracts, a van maker to discount or the grid operator to prioritize connections, a public commitment can change the odds of success and earns its cost. If it is a forecast dressed as a promise, it adds risk without changing the outcome.

Here the binding uncertainty, grid capacity, is outside Fernlea's control but will largely be resolved within two years. That favours B: it captures most of the credibility, keeps irreversible spending small until the key unknown is settled, and names the signposts that would justify a change, so a later revision looks like the plan working rather than failing. A is defensible only if the discount is large relative to the cost of being wrong, or if the firm order is what secures the grid connections. C is cheapest to reverse but gives away the coordination value customers want. Whichever is chosen, agree the review triggers now.

Section 08: Further reading

  • Frank H. Knight (1921). Risk, Uncertainty and Profit. Houghton Mifflin.

    The origin of the distinction between measurable risk and unmeasurable uncertainty. The opening chapters remain the clearest statement.

  • Daniel Kahneman and Amos Tversky (1979). Prospect theory: an analysis of decision under risk. Econometrica, 47(2).

    Prospect theory in its original form. Read for the reference point, loss aversion and risk-seeking in the domain of losses.

  • Amos Tversky and Daniel Kahneman (1974). Judgment under Uncertainty: Heuristics and Biases. Science, vol. 185, no. 4157, pp. 1124 to 1131.

    Short and foundational, on anchoring, availability and representativeness.

  • Daniel Kahneman and Gary Klein (2009). Conditions for intuitive expertise: a failure to disagree. American Psychologist, 64(6).

    Two rival schools agree on when expert intuition can be trusted. Essential for anyone who relies on experienced judgement.

  • Kathleen M. Eisenhardt (1989). Making fast strategic decisions in high-velocity environments. Academy of Management Journal, 32(3).

    A field study of how fast top teams decide well, overturning the idea that speed means less analysis.

  • Avinash K. Dixit and Robert S. Pindyck (1994). Investment under Uncertainty. Princeton University Press.

    The formal case for the value of waiting when investment is irreversible. Demanding, but the first two chapters are accessible.

Every work cited here is listed in the platform's sources register. Real organizations appear only with facts from the sources shown; everything else is labelled illustrative or fictional.