The operating condition: uncertainty is not a bug
Many institutional decision frameworks still assume a rational sequence: gather evidence, compare options, choose the best, implement. That model works reasonably well when objectives are clear, variables are measurable, causality is understood, and the environment does not change faster than the decision cycle. But the problems most institutions face today — governance reform, economic fragility, displacement, youth employment, complex crises — rarely meet these conditions.
They are closer to what Rittel and Webber called 'wicked problems': problems that are hard to define conclusively, have no definitive solution, and mutate as you intervene. In wicked settings, evidence does not converge neatly toward one correct answer. It often clarifies complexity rather than eliminating it. This is where institutions often misfire: they keep acting as if more evidence will turn uncertainty into certainty.
I. Evidence overload: when more knowledge produces paralysis
Information overload is not simply 'too much reading.' It is a predictable breakdown when the volume and complexity of information exceed what decision-makers can meaningfully process, especially under time pressure and accountability constraints. In institutions, overload has a particular shape: reports answer different questions at different time horizons using incompatible assumptions; findings are non-comparable or conflicting; uncertainty is hidden under confident language; and evidence arrives too late for the real decision window.
Between 2011 and 2020, the international development community collectively commissioned hundreds of governance assessments, political economy analyses, and institutional capacity studies on Tunisia. The country became one of the most-analysed political transitions of its generation. Yet the volume and contradictions among these assessments created a reform ecosystem in which decision-makers at ministerial level faced competing analytical frameworks and incoherent conditionality requirements. More analysis did not produce more clarity. It produced more defensibility for inaction.
Under these conditions, evidence fuels decision paralysis. Leaders sense that whatever they choose, there will be a document somewhere that could be used to criticize the choice. So the safest move becomes delay, additional analysis, or incrementalism. 'One more study' becomes institutionally rational: it postpones responsibility and reduces immediate exposure.
II. Bounded rationality: what actually replaces the rational ideal
Herbert Simon's foundational work on bounded rationality argues that decision-makers cannot evaluate all alternatives and consequences, so they settle for options good enough under real constraints rather than optimal. In practice, evidence is rarely used to identify the best choice. It is used to identify a choice that is defensible within the organization, feasible within available capacity, politically survivable, and procedurally clean.
Algeria's management of its subsidy system over three decades illustrates satisficing at institutional scale. Economists and international financial institutions consistently documented that Algeria's fuel, food, and utility subsidies were fiscally unsustainable and economically inefficient. The evidence was not disputed; it was acknowledged by technical actors within the government itself. Yet successive administrations satisficed: adjusting specific price points when oil revenues permitted, restoring subsidies when social unrest threatened, and commissioning further analysis to document constraints already well documented.
III. Cognitive constraints: why good evidence loses to simple stories
Under uncertainty, human judgment relies on heuristics — cognitive shortcuts that reduce complexity quickly but introduce systematic distortions. Tversky and Kahneman's foundational work identifies how people rely on representativeness, availability, and anchoring, producing predictable biases. In institutional life, salience beats significance: a vivid crisis dominates attention even when data suggests other priorities. Narratives beat nuance: evidence fitting a coherent story travels faster than evidence insisting on trade-offs.
Following the 2015 Bardo Museum and Sousse attacks, donor strategies pivoted dramatically. Funding streams supporting community-level prevention and structural drivers analysis were reoriented toward border security and law enforcement. The shift was driven primarily by salience: the vividness of the attacks overrode a more complex evidence base on radicalisation drivers, which pointed toward youth exclusion, governance grievances, and local social dynamics rather than security deficits. The evidence on what produced radicalisation had not changed. What changed was what decision-makers could see and feel.
IV. Power and politics: evidence competes with feasibility
The idea that evidence drives policy decisions as a neutral input is one of the most persistent institutional fantasies. Most high-level decisions allocate resources, create winners and losers, and shift power. Evidence may clarify consequences, but it cannot remove conflict. A more honest position is this: evidence influences decisions through political negotiation. The task is not to remove politics from evidence use, but to prevent politics from turning evidence into pure legitimation theatre.
Between 2015 and 2022, rigorous conflict analysis consistently found that externally brokered Libyan political frameworks lacked the domestic legitimacy, factional buy-in, and institutional architecture needed to hold. The international community proceeded anyway — not because it rejected the analysis, but because different external actors selectively cited whichever analytical findings supported their preferred political configuration. Evidence did not disappear. It was repurposed as legitimation material for decisions already taken on other grounds.
V. Audit culture: optimizing for defensibility, not outcomes
A major reason evidence fails to improve decisions is that many organizations now operate inside what Michael Power called an 'audit society': systems increasingly organized around verification rituals, internal controls, and demonstrable accountability. In audit-heavy environments, decisions are evaluated not only by whether they work, but by whether they are auditable. The safest decision is the one easiest to defend, not the one most likely to succeed.
Here is the central paradox: strong evidence can actually increase perceived liability. Evidence reveals complexity, uncertainty, and potential downside. If an institution is punished more for visible failure than for missed opportunity, ambiguity becomes dangerous — and the organization leans into standardized, low-variance choices that systematically underperform in complex realities.
VI. Learning is not deciding: why evaluations rarely change course
Carol Weiss's concept of 'knowledge creep and decision accretion' captures the mechanism: research influences policy gradually, by shaping the climate of ideas, rather than determining a specific decision at a specific moment. The result is a pattern most practitioners recognize: evaluations are produced, findings are acknowledged, decisions proceed with limited change. Learning systems and decision systems are misaligned: evidence is produced on 'learning timelines' while choices are made on 'political timelines' and 'budget timelines.'
Over the 2014–2021 period, multiple evaluations documented a consistent and deteriorating picture: declining public trust in parliament, mounting dissatisfaction with political parties, concentration of informal power, institutional gridlock, and erosion of constitutional governance culture. When President Kais Saied suspended parliament in July 2021, the evidence of institutional fragility had been publicly available for years. The problem was not analytical. It was architectural: no decision system existed that could translate accumulated evidence into protective action before the moment passed.
Implications for leaders who genuinely want decisions to improve
The practical question is not 'How do we produce more evidence?' It is: how do we make evidence usable in the real decision system we have — rather than the idealised one we wish we had. Three implications follow.
First, treat decisions as trade-offs, not as 'best practices.' In complex environments, evidence rarely prescribes. It clarifies consequences and risks. Institutions must make trade-offs explicit early, rather than hiding them under technical language that creates false consensus.
Second, separate facts, interpretation, and judgment — on purpose. If leaders cannot distinguish what is known, what is inferred, and what is a value choice, evidence will be weaponised or ignored. This is not a methodological nicety; it is a governance requirement.
Third, invest in decision support — not just research production. Decision support is the craft of turning an evidence environment into a choice environment: structuring options, clarifying risk layers, identifying feasibility constraints, and mapping second-order effects. It is not a summary. It is a governance function.
"Institutions make poor decisions despite good evidence because evidence is entering systems not designed to convert knowledge into judgment."