Decision-Making Under Uncertainty for Swiss Boards

Swiss corporate law asks directors to perform their duties with the care of an orderly and conscientious manager — a standard stated in Art. 717 OR and elaborated through the case law of the Swiss Federal Supreme Court. The standard is objective: the director is measured against the competent peer, not against their own private judgement of what reasonable care required. What the statute does not say — because it has no reason to — is that the humans subject to the standard decide under time pressure, with incomplete information, in rooms with other humans whose approval they want, about matters in which their own interests are often in play. The cognitive science of the last fifty years has documented, with some precision, the predictable ways in which human judgement under those conditions departs from the idealised director the statute presupposes. This article states the distance between the two, names the cognitive pressures that produce it, and argues that the procedural discipline Swiss corporate doctrine already prescribes — adequate information, conflict management, independent advice, contemporaneous documentation — is what closes the gap. Procedure is not the bureaucracy that decorates director duty. Procedure is the cognitive counter-measure on which director duty relies.

1. The objective standard and what it presupposes

Swiss doctrine under Art. 717 OR measures the director against the objective standard of the ordentlicher und gewissenhafter Geschäftsleiter — the orderly and conscientious manager. The standard is elastic: it adjusts to the company’s size, the decision’s stakes, and the sector’s specific demands. What it does not adjust to is the director’s private cognitive state. A director acting with what they subjectively believed was careful judgement can be liable if the objective standard was not met. The Swiss Federal Supreme Court has repeatedly affirmed this objective character; the business-judgement deference that Swiss courts extend is not a deference to subjective belief but to decisions reached through appropriate process — adequate information, absence of conflict, orientation to the company’s interest, and reasonable deliberation. The objective-standard architecture is developed in BGE 113 II 52 and refined in BGE 139 III 24, with procedural-deference contours elaborated in subsequent jurisprudence.

The standard presupposes a director who, given adequate information and an unconflicted position, will identify the company’s interest and decide accordingly. The director is assumed to weigh alternatives on their merits, to assess probabilities in proportion to evidence, to update prior views when new information arrives, and to distinguish the favoured outcome from the correct one. These are strong assumptions. The cognitive research of the last half-century shows them to be, at best, aspirational: human decision-making — even the decision-making of trained, experienced, and careful professionals — departs from the ideal in systematic and predictable ways.

This is not a doctrinal problem. It is a reason for the procedural architecture that surrounds the substantive standard. The Swiss business-judgement deference operates precisely on process: decisions reached through proper information, absent conflict, with reasonable deliberation, are reviewed deferentially. Decisions reached through thin process are reviewed directly. Whether or not Swiss doctrine developed with the cognitive-science literature in mind — and the two strands developed largely in parallel — it aligns with the insight that substantive judgement is fragile and procedural conditions are what make it defensible.

2. Cognitive hazards at the board table

Kahneman’s synthesis of the field describes human cognition as operating through two systems: a fast, intuitive, effortless system that produces most of our judgements, and a slower, deliberate, effortful system that can override the first when engaged (Kahneman, 2011). Board decisions are ordinarily the province of System 2 — directors know they are deciding consequentially and intend to think carefully — but the System 1 tendencies remain active, and several of them show up reliably in board settings.

Anchoring. The first number put in front of a committee becomes the reference against which later numbers are judged, even where the anchor is known to be arbitrary (Tversky & Kahneman, 1974). In board practice: the executive’s proposed compensation figure sets the band within which the committee debates; the opening settlement offer sets the range within which the board negotiates; a dividend equal to last year’s feels like the default even where last year’s circumstances have changed.

Availability. Events that come readily to mind are judged more probable than events that do not (Tversky & Kahneman, 1973). Boards reviewing risk attend disproportionately to the risk that was in the last briefing or in the last commentary piece; risks that the company has never experienced are under-weighted regardless of their actual base rate.

Loss aversion and the endowment effect. Losses are felt roughly twice as strongly as equivalent gains (Kahneman & Tversky, 1979). Directors resist decisions that crystallise losses — writing off an investment, dismissing an incumbent executive, pursuing a known claim — even where the counter-factual (continued drift) is worse. The endowment effect makes incumbent arrangements — an existing CEO, an existing compensation structure, an existing auditor — feel more valuable than they would to a fresh evaluator.

Over-confidence. Humans — and notably experts — often over-estimate the accuracy of their own probability judgements (Fischhoff, Slovic & Lichtenstein, 1977). The expert-calibration literature is not uniformly bleak: in domains with tight, frequent feedback — weather forecasting, competitive bridge, tournament chess — experts are substantially better calibrated than novices. The effect is severe, however, in domains where feedback is slow, noisy, or arrives too late to discipline the next judgement. Tetlock’s longitudinal work on political forecasters (Tetlock, 2005) is the clearest demonstration in that second category; much board-level judgement about strategy, litigation exposure, and executive performance sits in the same class. Directors asked to assess their own company’s litigation exposure, their own board’s oversight, or their own executive team’s competence are systematically optimistic — not because they are unusually prideful but because the feedback loops are long, the counter-factual is invisible, and their own conduct is what they see most clearly and most sympathetically.

Motivated reasoning. Humans think differently about propositions depending on whether they want them to be true (Kunda, 1990). A director who wants a transaction to go ahead sees the evidence for it more clearly and the evidence against it less clearly than a neutral observer would. The sensation is of reasoning carefully; the output is skewed toward the preferred conclusion. Conflict-of-interest doctrine exists because the law has no way to check the sensation.

Group-level biases. Boards are groups, and groups produce their own effects beyond those of their members. Janis’s Groupthink (1972) catalogued a consistent pattern: cohesive groups under pressure converge on a preferred conclusion, marginalise dissenting views, rationalise setbacks, and develop an illusion of consensus that persists despite private doubts. Asch’s conformity experiments (1951) demonstrated that even on visually obvious judgements, individuals systematically suppress their own perception to match the stated view of the group.

Escalation of commitment. Once a group has publicly committed to a course, subsequent evidence against it tends to reinforce rather than dislodge the commitment (Staw, 1976). Each prior step raises the sunk cost of reversal, and the reversal itself is felt as a loss. Boards in active M&A processes, boards persisting in failing strategic initiatives, and boards defending litigation well past the point of recoverable value all exhibit this pattern.

Automation bias. Humans over-trust outputs produced by machines, particularly where the machine’s confidence exceeds the human’s own (Parasuraman & Riley, 1997). As AI systems intermediate board-facing outputs — risk scores, monitoring summaries, generated memoranda — this bias becomes a direct oversight concern. See the comparative commentary on Marchand v. Barnhill and the reference article on AI governance.

Fundamental attribution error. When explaining others' behaviour, humans systematically under-weight situational causes and over-weight dispositional ones (Ross, 1977). Boards reading an executive failure as “wrong person for the job” rather than “wrong structure around the person” commit this error at scale, and respond by changing the person rather than the structure. The attribution matters not only for fairness to the departed executive but for whether the next incumbent inherits the same failure conditions.

Probability neglect. Small probabilities are not treated proportionally when the possible outcome is affectively loaded; a one-per-cent chance of catastrophe and a five-per-cent chance of catastrophe can be cognitively indistinguishable to a board whose attention is fixed on the severity rather than the frequency (Sunstein, 2002). The pattern cuts both directions: tail-risk oversight tends either to over-react to a single salient recent event or to neglect a known low-probability exposure, depending on which currently has attention. A disciplined reserve analysis forces the question the intuition would otherwise skip: not “is the catastrophe possible” but “how large is the product of probability and magnitude.”

Status-quo bias. Distinct from loss aversion: humans prefer the present arrangement to an equivalent alternative simply because it is the present arrangement (Samuelson & Zeckhauser, 1988). At board level this appears as the persistent continuation of arrangements — an incumbent CEO, an incumbent auditor, an incumbent group structure, an incumbent dividend policy — whose continuation would not be chosen if the board were setting up the company fresh. The disciplined test is the reverse-reverse question: if we did not have this arrangement today, would we create it?

Hindsight bias. After an outcome is known, humans systematically misremember their prior uncertainty — the outcome seems more predictable ex post than it was ex ante (Fischhoff, 1975). This affects both the director reflecting on past decisions (they believe they saw the risk coming, even where they did not) and the court reviewing those decisions (a judge asked whether a competent director would have foreseen the outcome will tend to believe the answer is yes). The Swiss reliance on contemporaneous documentation exists partly because of this: the record made before the outcome is known is the corrective to the reconstruction made after.

3. Disciplines that counter the hazards

Swiss corporate practice already carries, in its procedural architecture, several disciplines that counter the cognitive pressures above. Each can be read as a cognitive counter-measure as well as a legal requirement; reading them that way sharpens why they matter.

Information pre-distribution. Board papers sent twenty-four or forty-eight hours in advance allow directors to engage System 2 before the meeting rather than during it. Papers distributed in the room produce predictable under-analysis; under time pressure, directors defer to the paper’s framing and to the executive’s oral summary. A board that routinely receives materials late should read that not as an operational annoyance but as a structural cognitive failure.

Independent advice. The Swiss reliance-on-competent-advice doctrine — a board may discharge its duty of care by relying on advice that is independent, properly instructed, and duly weighed — is a direct response to expert over-confidence. An independent adviser does not have the director’s motivated reasoning about the decision because the adviser’s future does not depend on the decision’s outcome. The protection the doctrine offers depends precisely on the adviser being independent; advice that is procured to ratify a conclusion already reached does not defeat motivated reasoning, it implements it.

Structured dissent. Janis’s prescription for groupthink included formal devil’s advocacy: a named member assigned to argue against the proposal on its merits, rotated across meetings so that the role does not attach to a single dissenter. Klein’s pre-mortem technique (Klein, 2007) has the group imagine the decision has already failed and work backwards to identify the causes — a method that bypasses the motivated-reasoning resistance that surfaces when the same question is asked in forward form. The reference-class forecasting technique (Kahneman & Tversky, later Flyvbjerg) forces the board to find the base rate of comparable past decisions rather than anchor on the present decision’s internal logic alone. All three are available to Swiss boards today; see the prompts library for structured versions.

Cognitive diversity. Boards whose members share backgrounds, training, and network origins are more vulnerable to availability and groupthink; boards whose members bring different prior frames are more likely to surface alternative readings before a decision hardens. Cognitive diversity is not the same as demographic diversity, though the two correlate. The Swiss compensation-committee independence requirements under Art. 733 OR address conflict-based cognitive bias; broader cognitive diversity on the full board addresses the subtler forms.

Conflict management. Motivated reasoning is strongest where the director has something at stake. Art. 717a OR’s disclosure-and-recusal discipline, and the conflict-free-committee pattern for material transactions, exist because the cognitive damage of divided loyalty cannot be repaired by will-power alone. The director who believes they can “be fair about this” is exhibiting the bias, not defeating it.

Documentation. Minutes written contemporaneously, capturing the deliberation rather than only the result, are the corrective to hindsight bias. A decision later reviewed with knowledge of the outcome is rated better or worse than it deserved; the contemporaneous record preserves the information actually available at the time and the reasoning actually applied to it. This is what the Swiss courts look for first in a director-liability case. See the Litigation Readiness article for the discipline at length.

4. The architecture of the board meeting

Cognitive quality is widely thought to degrade across the length of a meeting. Baumeister’s ego-depletion research (Baumeister et al., 1998) and Danziger’s study of judicial parole decisions over a working day (Danziger et al., 2011) have been influential accounts of the mechanism; both have been substantially contested in the replication literature — large-scale multi-lab studies have not reproduced the classical ego-depletion effect (Hagger et al., 2016), and the judicial-fatigue pattern has been attributed in reanalysis to case-ordering selection rather than depletion (Weinshall-Margel & Shapard, 2011). The practical intuition — that consequential decisions taken late in long meetings are reached on depleted attention — remains a prudent default, even where the specific findings are less settled than the popular accounts suggest. The implication for board agendas is straightforward: the most consequential decision is placed early, because the architecture of the agenda shapes the deliberation rather than containing it neutrally. The consent-agenda pattern that consolidates routine approvals frees deliberate attention for the matter where it is needed.

Sequencing. The order of items affects both attention and anchoring. A board discussion of a question that has been preceded by a twenty-minute management presentation on an adjacent topic is a discussion already anchored; boards that wish the framing to be their own rather than management’s sometimes hold the sensitive item before management presents.

Duration. Meetings too long to be productive are a common pathology; meetings too short to be deliberate are worse. A sixty-minute meeting covering eight items cannot decide any of them well; directors who accept such agendas are accepting the decision-quality consequences. The auditor- only session and the conflict-free-committee session, each given its own time, are responses to the cognitive limits of combined deliberation.

Physical and structural conditions. The seating around the table, the acoustic quality of the room, the presence or absence of management in particular deliberations — each shapes who speaks and what is said. These are not secondary to the decision; they are conditions on it. A disciplined chair attends to them.

5. When heuristics serve the board

Not every decision benefits from formal analysis. Gigerenzer’s research on ecological rationality argues that simple heuristics — “take the best,” “satisficing,” recognition- primed decision — outperform elaborate analysis in environments where information is noisy and the decision must be reached fast (Gigerenzer & Selten, 2001; Klein, 1998). Experienced directors, like experienced clinicians and firefighters, develop pattern-recognition that produces fast high-quality judgement in familiar domains. The board that treats every decision as requiring a formal reference-class forecast will exhaust itself; the board that treats no decision as requiring one will mis-calibrate under the biases catalogued above.

The practical discipline is a calibrated split. Routine operational items — the consent agenda — benefit from heuristic speed. Novel, high-consequence, or emotionally charged decisions — distressed-balance-sheet choices, executive dismissals, crisis responses, major M&A, AI deployments — benefit from deliberate structure. The test is not the dollar value but the conditions under which the bias would operate. Familiar decisions with short feedback loops are safe heuristic territory; unfamiliar decisions with long feedback loops require the procedural counter- measures.

The Swiss business-judgement doctrine accommodates this calibrated split. Deference to decisions reached through adequate information and reasonable deliberation does not mean formal analysis of every item; it means appropriate analysis of each item, with the appropriateness calibrated to the stakes.

A rigorous reader should press on the implicit claim that procedural discipline is what distinguishes good boards from mediocre ones. A plausible opposing view — held, in the author’s practice experience, by a number of senior directors — is that what distinguishes first-rate boards is not procedure but substantive judgement: the pattern recognition that comes from sustained engagement with the sector, the instinct that flags a problematic deal before the structured analysis has reached it, the domain expertise that reads a balance sheet or a strategic pivot at sight. On this view, procedure is the floor any competent board meets; it is not the ceiling, and it is not what separates excellent boards from adequate ones.

This is a serious objection and the honest answer is that both claims are partly right. Procedural discipline is necessary but not sufficient: a board that meets the procedural gates but lacks substantive judgement will misread the strategic situation, and the record — however well-minuted — will record a careful process toward a poor outcome. Substantive judgement is necessary but not sufficient either: a board of excellent judges that operates without structured information, without conflict management, and without contemporaneous documentation will see its judgement distorted by the cognitive pressures catalogued in §2 and will not leave the record it needs in litigation. The two are complements, not substitutes.

The position this article takes is therefore modest. Procedure is the part of board quality that can be deliberately cultivated, measured, taught, and reviewed under Art. 717 OR. Substantive judgement — while arguably the more decisive variable at the margin — is not. A reference work of this kind can speak usefully about the first and should be humble about what it can say about the second. What it can say about the second is that procedure makes space for judgement to operate: a conflict-free committee frees the experienced director to reason without motivated reasoning; pre-distributed papers free working memory for pattern recognition; independent advice puts substantive judgement in dialogue with a second substantive judgement rather than with its own echo. In this sense the procedural architecture is not a substitute for judgement. It is the condition under which judgement becomes visible — to the board itself, to the shareholders, and to the court that may later review the decision.

6. What this means for boards

Three practical points follow.

The objective director standard is met through procedural discipline. A director cannot improve their cognitive hardware; they can only improve the conditions under which it operates. Adequate information in time, unconflicted deliberation, independent advice on specialist questions, contemporaneous documentation, and structured dissent on high-stakes items — these are not procedural decoration. They are the conditions under which human judgement reaches the standard the law requires.

Procedure is the cognitive counter-measure that matches the cognitive hazard. Not every procedural element addresses every bias. The written reserve analysis addresses optimism about contingent exposures; the conflict-free committee addresses motivated reasoning; the auditor-only session addresses authority deference; the pre-mortem addresses confirmation bias; the independent external adviser addresses expert over-confidence; the contemporaneous minute addresses hindsight distortion. Matching the procedure to the hazard is the work.

Self-awareness is part of the discipline. A director who knows that loss aversion will bias them against writing off a bad investment is better positioned to resist the bias than one who does not; a committee whose members understand that their benchmark will anchor their recommendation can be more sceptical of the benchmark. The cognitive-science frame is not an excuse for bad decisions — it is the opposite. It is the account of why the procedural discipline matters, and of what fails when the discipline is thin.