Knee-Deep in the Mud and Climbing Out with a Clean Face

A History-Aware Multiobjective View of Escalation and De-escalation of Commitment

We usually expect people to change course after negative feedback. Yet in investment and project settings the opposite is common: decision makers sometimes increase their commitment to a failing course of action. This pattern is known as escalation of commitment (Staw 1976; Staw 1981; Brockner 1992).

The classic empirical anchor is Staw’s study Knee-Deep in the Big Muddy (Staw 1976). In a simulated business investment task, commitment rose most when individuals were personally responsible for earlier choices that produced negative consequences. The interpretation emphasized self-justification: continuing can protect identity and credibility even when outcomes worsen.

Staw’s later review (Staw 1981) and Brockner’s synthesis (Brockner 1992) helped establish escalation as a multi-causal phenomenon involving responsibility, face-saving, organizational pressures, and the framing of losses and alternatives.

A closely related stream extends the idea from investments to escalation of conflict and “entrapment,” notably in the book by Brockner and Rubin (Brockner & Rubin 1985).

De-escalation is not just “doing less”

The literature also recognizes de-escalation of commitment: deliberately reducing further investment, changing direction, or terminating a course of action. Importantly, de-escalation can itself be shaped by cognitive and social mechanisms rather than being a simple “rational reset” (Heath 1995; Molden & Hui 2011; Moser, Wolff & Kraft 2013).

Heath (1995) explicitly analyzed both escalation and de-escalation in response to sunk costs, proposing that mental budgeting can drive either continued investment or withdrawal depending on how expenses are tracked and bounded.

More recent work suggests concrete de-escalation levers, such as predecisional accountability (Moser, Wolff & Kraft 2013) and regulatory-focus framing that can loosen self-justification motives (Molden & Hui 2011).

Why reversals feel uniquely painful

Here is the psychological nuance that matters for interactive decision support: a direct reversal of earlier choices can look like obvious wasted effort. It is highly visible, invites scrutiny, and puts the decision maker in a bad light—especially under personal responsibility, which is exactly the escalation trigger identified in early work (Staw 1976; Brockner 1992).

In contrast, repair moves can be narrated as prudent adaptation: “we learned,” “we improved robustness,” “we mitigated risk.” The resource cost may be comparable, but the social meaning is different. This helps explain why some teams prefer to add layers of fixes rather than admit and undo earlier commitments (Staw 1981; Heath 1995).

Irrationality, state-dependence, and the “small steps become big steps” effect

Escalation research sits inside a broader family of findings on bounded rationality and systematic bias. A particularly influential anchor is prospect theory. Kahneman and Tversky argued that people evaluate outcomes relative to a reference point and that losses loom larger than gains of the same magnitude (Kahneman & Tversky 1979; Tversky & Kahneman 1992). This makes the current state of an investment or project psychologically crucial: once the situation is framed as a loss, decision makers may become risk-seeking in the hope of “getting back to even.”

A second, more subtle ingredient is that small preference changes can accumulate into a clearly visible big change. Locally, a decision maker may treat each small step as “almost the same,” but the total displacement can become clearly non-negligible. This is related to classic discussions of intransitivity and the limits of indifference when judgments are made in small increments (Tversky 1969). In practice, this can make long sequences of tiny adjustments feel safe and justifiable, even if the overall trajectory has drifted far from the original intent.

These two ideas already suggest that a purely state-based model may be incomplete. In the framework below, we therefore go one step further: we incorporate not only the current state, but also the path or history of moves that led to it. This turns explanation and accountability into part of the model itself.

Connecting this to interactive multiobjective optimization

Interactive multicriteria optimization (IMO) is built on iterative preference articulation and course correction. Escalation of commitment highlights a hidden barrier: the process cost of admitting error can distort preference updates, making the decision maker reluctant to move reference points away from earlier choices (Staw 1981; Brockner 1992).

To model this realistically, it is not enough to look only at the current decision state. We also need to consider how we got here. In other words, the decision support problem depends on:

the current state, and
the sequence of earlier moves that created it, including partially bad choices.

A history-aware “bag of moves” model

Let a decision process start from a baseline ${x_0}$ and apply a sequence of ${k}$ moves with changes ${\Delta_1,\dots,\Delta_k}$ :

${x_k = x_0 + \sum_{i=1}^k \Delta_i.}$

Let ${P(x)}$ denote the final outcome of interest, for example profit (larger is better). We can evaluate the past moves using a Shapley-value idea. Define:

${g(S) = P\!\left(x_0 + \sum_{i\in S}\Delta_i\right)}$

for any subset ${S}$ of the historical moves. The Shapley value of move ${i}$ , call it ${\phi_i}$ , gives a principled estimate of how much that move contributed to the final profit on average across all combinations of moves.

Interpretation:

${\phi_i > 0}$ : the move likely helped profit.
${\phi_i < 0}$ : the move was likely counterproductive.

This gives a structured way to discuss “wasted effort” in a path-dependent sense: not only where we are, but which earlier steps helped or harmed the outcome.

Two kinds of de-escalation actions

Now consider a deliberate exit from a potentially failing course of action. We distinguish two kinds of actions:

Repair moves ${R}$ : new positive steps that mitigate damage without explicitly admitting earlier mistakes.
Undo moves ${U}$ : explicit reversals of some earlier moves, which can publicly reveal that effort was wasted.

A simple way to express the resulting state is:

${x' = x_0 + \sum_{i \notin U} \Delta_i + \sum_{j \in R} \Delta^{R}_j.}$

Here ${U}$ is a subset of the earlier (historical) moves, and ${R}$ is a set of newly chosen repair moves.

Costs that feel different

Let each repair move ${j}$ have an operational cost ${c^{R}_j}$ . These costs are real, but they can often be framed as sensible adaptation.

Undo moves have a different kind of price. Let each earlier move ${i}$ have an “exposure cost” ${e_i}$ , and (optionally) a visibility factor ${v_i}$ with ${0 \le v_i \le 1}$ . The total cost of exposing past mistakes through undoing is:

${C_{\text{expose}}(U) = \sum_{i \in U} v_i\, e_i.}$

This captures your key psychological point: reversing an earlier choice can be socially salient in a way that adding repairs is not.

The bi-objective Pareto view

This leads to a compact multiobjective formulation in which the decision maker balances:

Final outcome (e.g., profit after repairs and undoing), and
Cost of exposing past mistakes.

Formally, we can view the de-escalation decision as:

maximize ${P(x')}$
minimize ${C_{\text{expose}}(U)}$

The Pareto front then shows a spectrum of strategies ranging from:

high-performance turnarounds that require explicit reversals, to
face-preserving paths that rely more on repairs and fewer visible undo steps.

The Shapley values ${\phi_i}$ can guide this exploration: if some earlier moves have strongly negative ${\phi_i}$ , they become candidates for undoing—yet each such undo can increase ${C_{\text{expose}}}$ . That is the core tension.

This is also what makes the model feel new: unlike many earlier formulations that focus on the present state alone, this one is history-aware. It explicitly treats the path of prior (partially bad) decisions as an object of evaluation and optimization.

Why this is useful

Seen this way, escalation is not just “irrationality.” It can be interpreted as an implicit multiobjective problem in which decision makers trade off outcome quality against the social costs of visible reversal. Making these process criteria explicit in IMO could support more realistic, humane, and ultimately more effective de-escalation pathways (Heath 1995; Molden & Hui 2011; Moser, Wolff & Kraft 2013). In the end these ideas should be seen as a thought experiment rather than actionable advice.

The best de-escalation in terms of final outcome, in the end, is achieved when all actions are considered—and when the surrounding culture makes that possible. In particular, de-escalation becomes easier and more effective in environments where mistakes are not treated as failures but as learning processes and experience gain, and where honest reflection on one’s past mistakes is seen as a positive character and leadership trait.

References

Staw, B. M. (1976). Knee-deep in the big muddy: A study of escalating commitment to a chosen course of action. Organizational Behavior and Human Performance, 16(1), 27–44.
Staw, B. M. (1981). The escalation of commitment to a course of action. Academy of Management Review, 6(4), 577–587.
Brockner, J. (1992). The escalation of commitment to a failing course of action: Toward theoretical progress. Academy of Management Review, 17(1), 39–61.
Brockner, J., & Rubin, J. Z. (1985). Entrapment in Escalating Conflicts: A Social Psychological Analysis. New York: Springer-Verlag.
Heath, C. (1995). Escalation and de-escalation of commitment in response to sunk costs: The role of budgeting in mental accounting. Organizational Behavior and Human Decision Processes, 62(1), 38–54.
Molden, D. C., & Hui, C. M. (2011). Promoting de-escalation of commitment: A regulatory-focus perspective on sunk costs. Psychological Science, 22(1), 8–12.
Moser, K., Wolff, H.-G., & Kraft, A. (2013). The de-escalation of commitment: Predecisional accountability and cognitive processes. Journal of Applied Social Psychology, 43, 363–376.
Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–291.
Tversky, A. (1969). Intransitivity of preferences. Psychological Review, 76(1), 31–48.
Tversky, A., & Kahneman, D. (1992). Advances in prospect theory: Cumulative representation of uncertainty. Journal of Risk and Uncertainty, 5, 297–323.

Please cite as:
Michael Emmerich (2025, December 9). Knee-Deep in the Mud and Climbing Out with a Clean Face: A History-Aware Multiobjective View of Escalation and De-escalation of Commitment. MODA news (www.emmerix.net).