Predictive leadership: minimising the metabolic tax of prediction error for sustainable performance Open Access

Leadership in contemporary organisations extends far beyond operational management; it requires the capacity to navigate complexity, foster trust, and sustain psychological safety in fast-changing environments (Edmondson and Bransby, 2023; Salameh-Ayanian et al., 2025; Westover, 2024). For decades, practitioners have relied on established models—such as Situational, Transformational, Servant, and Authentic Leadership—to guide these efforts (McCleskey, 2014; Pock and Pangaro, 2023; Shi and Zhou, 2023). These frameworks have served as essential scaffolds, providing leaders with a shared language and structured approaches to adapt their style to the context, inspire vision, and ethically serve their teams (Butko et al., 2025; Mahon, 2020; Soelistya et al., 2024).

However, organisations continue to report persistent difficulties with trust, psychological safety, and change adoption, even when these models are applied (Edmondson and Bransby, 2023; Mahon, 2020). We argue that this persistence stems from a fundamental limitation: traditional leadership models—whether Situational, Transformational, or Servant—are primarily reactive (Ejaz et al., 2025; Hansbrough and Schyns, 2018). They operate on a behavioural logic that asks leaders to respond to what followers do in the moment (Boyar et al., 2023; McCleskey, 2014). This reactive stance is insufficient for managing modern psychosocial risks (Deng et al., 2023; Westover, 2024).

We posit that leadership must instead function as a primary upstream control (Butko et al., 2025; Gendron and Barrett, 2019). By stabilising high-stakes interactions, such as how decisions are communicated and how change is framed, leaders do not just influence sentiment; they reduce the “hazard exposure” of prediction error (Clark, 2013; Ohira, 2018). This shifts leadership from a soft skill of influence to a structural mechanism for hazard prevention. Crucially, because these predictions are constructed from past experiences and conceptual knowledge, the framework posits that effective leaders must actively manage the specific language and memory cues that followers use to simulate the future. Current organisational responses to psychosocial risk often rely on downstream, reactive measures such as resilience training or wellness support that attempt to treat the symptoms of stress without altering its cause (Dhingra and Punia, 2016; Tan, 2024).

Modern neuroscience reveals that the human brain is not a reactive organ, but a predictive one (Clark, 2013; Ohira, 2018). Established frameworks like Dual Process Theory explain what is occurring: the brain relies on “System 1” thinking—fast, automatic, and associative processing—to navigate the world efficiently (Kahneman, 2011). Simultaneously, research on Implicit Leadership Theories (ILTs) confirms that followers rely on preconscious cognitive categories, or “prototypes,” to interpret leader behaviour automatically (Lord et al., 2020; Shondrick and Lord, 2010).

While these theories describe the cognitive mechanics, recent advances in Constructed Emotion Theory (CET) explain the biological why behind them (Feldman-Barrett, 2017; Gendron and Barrett, 2019). Drawing on the principle of allostasis, we posit that the brain functions as a prediction engine designed to manage the body's energy resources efficiently (Feldman-Barrett, 2017). Prediction is a physiological preparation for action; the brain anticipates sensory inputs to prepare the body for metabolic expenditures before they arise (Hoemann et al., 2019b; Ohira, 2018). When a leader's actions violate these internal predictions, the result is not merely cognitive confusion, but a “metabolic tax” (Ohira, 2018). The brain must mobilise expensive resources to process the prediction error, triggering a physiological stress response often experienced as dissonance or resistance (Lopez-Valeiras et al., 2022).

Crucially, while Kahneman (2011), Lord et al. (2020), and Feldman-Barrett (2017) provide the descriptive map of what these processes are and why they exist, they do not offer leaders the prescriptive tools to manage them in an organisational setting. Knowing that resistance is a metabolic cost does not tell a leader how to prevent it.

To operationalise prediction management, we must first distinguish the descriptive mechanisms of social cognition (how we think) from the mechanistic drivers of physiology (why we think).

1. The Descriptive Landscape: System 1 and Implicit Leadership Theories

Human cognition relies on efficiency. As established in Dual Process Theory, the brain operates primarily via “System 1”—a fast, automatic, and associative mode of processing that prioritises speed over accuracy (Kahneman, 2011). This aligns with what organisational psychology terms Implicit Leadership Theories (ILTs) (Lord et al., 2020; Shondrick and Lord, 2010).

Prediction is grounded in the brain's ability to recognise patterns, drawing on stored experiences to form mental models of likely outcomes (Clark, 2013; Lord et al., 2020). These mental models act as filters for perception, shaping behaviours and emotional responses before observable actions occur (Gendron and Barrett, 2019; Lord et al., 2020; Lord and Maher, 1991). In organisational life, employees develop anticipatory frameworks: “What happened last time?”, “What is likely to happen if I speak up?”, “What has this leader done before?” (Boyar et al., 2023; Lord et al., 2020). When a leader's behaviour matches these internal prototypes (e.g. “Tyrant” or “Hero”), recognition is fast, and trust is automatic (Shi and Zhou, 2023). However, when behaviour violates the prototype, it triggers dissonance (Lopez-Valeiras et al., 2022).

1. The Mechanistic Core: Prediction as Allostasis

Why do these prototypes exist? Modern neuroscience confirms that they are not merely cognitive shortcuts, but tools for energy regulation (Feldman-Barrett, 2017; Ohira, 2018). The brain actively constructs expectations from prior experiences, using prediction to navigate both present environments and future possibilities (Clark, 2013; Hoemann et al., 2019b). Drawing on the work of Feldman-Barrett (2017) and Sterling (2012), we posit that the brain's primary function is allostasis. Feldman-Barrett (2017) refers to this simply as managing a “Body Budget”: the brain runs an internal model of the world to anticipate and balance metabolic resources—such as glucose, oxygen, and salt—just as a finance department balances cash flow.

Constructed Emotion Theory (CET) reinforces this (Feldman-Barrett, 2017). CET proposes that emotions are not innate, universal reactions but are constructed through predictive processes shaped by cultural norms, language, prior experiences, and emotional concepts (Butko et al., 2025; Liu et al., 2024). Emotions function as predictive simulations, preparing the body and mind for anticipated events rather than merely responding after the fact (Brooks et al., 2017; Hoemann et al., 2019b). Therefore, a “prototype” is not just a mental image; it is a pre-packaged metabolic budget (Feldman-Barrett, 2017). Following the prototype is metabolically cheap; processing new, unexpected information is expensive (Clark, 2013; Ohira, 2018).

While a single prediction error incurs a momentary metabolic cost, chronic unpredictability creates a compounding deficit (Feldman-Barrett, 2017; Ohira, 2018). When a leader's behaviour remains unpredictable, the follower's brain cannot “switch off” the error signal, even outside of work hours (Sahoo et al., 2023). This creates a recovery deficit where the metabolic cost of work spills over into personal time, impairing sleep and physiological restoration (Kjærgaard et al., 2024; Levitats et al., 2022). The employee returns to work the next day in a state of re-entry drag, starting the day with reduced cognitive bandwidth. Crucially, reactive leaders often misdiagnose this biological depletion as a performance or attitudinal issue, applying further pressure that accelerates the cycle of harm and reinforces protective beliefs and predictions (Daniels and Robinson, 2019; Levitats et al., 2022).

1. Prediction Error as a “Metabolic Tax”

This biological lens reframes the nature of organisational friction. Prediction error occurs when the brain's expectations do not align with reality—when anticipated outcomes fail to materialise or when unexpected outcomes arise (Clark, 2013; Ohira, 2018). When this happens, the brain cannot rely on its efficient System 1 model. It is forced to recruit expensive neural resources to process the novelty and update the model (Feldman-Barrett, 2017; Kahneman, 2011).

We argue that this prediction error imposes a “metabolic tax” on the employee (Feldman-Barrett, 2017; Ohira, 2018). The mismatch creates a state of physiological arousal often experienced as cognitive dissonance (Lopez-Valeiras et al., 2022). If unresolved, this metabolic cost accumulates (Levitats et al., 2022). Leaders who fail to recognise and address prediction error risk leaving employees in unresolved dissonance, which can undermine trust, engagement, and performance (Iddrisu, 2025; Joo et al., 2023).

Unresolved prediction error triggers a physiological stress response—mobilising cortisol and glucose—to deal with the unknown (Feldman-Barrett, 2017; Ohira, 2018). If this error persists, the brain attempts to balance its energy budget by withdrawing resources, often manifesting as disengagement or fatigue (Barrett et al., 2016). Resistance to change, therefore, should be reframed not as a behavioural choice, but as a protective metabolic response to the high energy cost of updating deep-seated predictions (Clark, 2013).

Together, insights from neuroscience and CET show that behaviour is shaped in advance through predictive mechanisms. Leadership models that do not engage with these upstream processes miss the opportunity to influence future behaviour, tending to rely on reactive strategies that address issues only after the metabolic cost has already been incurred.

To avoid this metabolic tax, the brain often engages in data sorting, ignoring contradictory information to preserve its existing prediction models. In this context, bias is reframed not as a moral failing but as a “stubborn prediction”—a resource-conservation strategy the brain uses to avoid the bio-energetic cost of learning new patterns (Feldman-Barrett, 2017). As Kahneman (2011) establishes, System 1 seeks coherence over accuracy. In an organisational context, the brain reduces uncertainty by forcing new data to fit old patterns. Predictive Leadership, therefore, views bias as a physiological refusal to incur the cost of updating deep-seated models (Lopez-Valeiras et al., 2022; Russen et al., 2025).

While Dual Process Theory and Implicit Leadership Theory provide the essential descriptive map of these cognitive and physiological landscapes, they do not provide the prescriptive tools to navigate them. Knowing that followers hold implicit prototypes (Lord et al., 2020; Shondrick and Lord, 2010) or that the brain budgets energy (Feldman-Barrett, 2017) does not tell a leader what to do during a merger or a crisis. To visualise this shift, Figure 1 illustrates the full “Predictive Leadership Process.” It maps the flow from upstream inputs—specifically the stable “Geology” of past experience and the volatile “Weather” of the current state—through the brain's prediction engine (Clark, 2013; Ohira, 2018). Crucially, it highlights the bifurcation at the “Reality Check” (Zone 3): matches produce “Metabolic Ease” (trust and flow), while mismatches trigger a “Metabolic Tax” (stress and resistance). This mechanistic view explains why the specific leadership interventions outlined in Table I  are necessary to regulate the team's energy economy.

Predictive Leadership fills this gap. It operationalises the neuroscience of prediction into a practical management system. It shifts the leader's focus from managing downstream consequences (behaviour) to architecting the upstream antecedents (expectations and physiological safety) (Edmondson and Bransby, 2023; Iddrisu, 2025). Ultimately, the goal is to identify the prediction upstream—prior to the event—to validate the follower's reality and architect a positive outcome that minimises the metabolic tax of uncertainty.

To operationalise this framework, we distinguish between two specific time horizons of prediction: the deep, stable history of the group (“The Geology”) and the volatile, immediate context (“The Weather”). This distinction prevents the redundancy often found in leadership models by separating the management of long-term memory structures from the management of immediate physiological states. The framework then moves to the mechanics of prediction itself: detecting the error (“The Diagnostic”) and constructing new models (“The Architecture”).

Table 1 summarises observable behaviours; the subsections that follow outline the rationale and applications.

1. Mapping Past Experience (The Geology)

This practice addresses the deep, stable cognitive structures identified by Lord et al. (2020) as Implicit Leadership Theories (ILTs) (Gershman, 2017; Lord et al., 2020). Just as geology dictates where water flows, an organisation's history and the “prototypes” stored in followers' long-term memory dictate how current leadership actions are interpreted (Gershman, 2017; Zacks et al., 2022). These prototypes are not observable behaviours; they are internal, calcified expectations formed over years of experience (Hansbrough and Schyns, 2018).

Leaders often fail to account for these “ghosts”—previous leaders, past traumas, or deep-seated cultural myths. Neuroscience suggests these are not just memories; episodic experiences consolidate over time into semantic schemas (prototypes) that function as pre-packaged metabolic budgets (Feldman-Barrett, 2017; Lord et al., 2020). The brain relies on these stable prototypes because they are metabolically cheap to run and help plan for future action (fight or flight) (Clark, 2013; Ohira, 2018). When a new leader enters a team with a history of toxic management, the team's brain defaults to the “Tyrant” prototype, not out of malice, but out of bio-energetic conservation (Shi and Zhou, 2023; Shih and Yeh, 2024). It is less expensive to predict a known threat than to process the high-metabolic cost of learning a new leadership style. Therefore, the team is predicting “threat” based on their geology to conserve energy and prepare for future protective behaviours.

Instead of asking generic questions about expectations, leaders must actively surface the historical data followers are using to simulate the future. This involves specific lines of inquiry such as:

1. “Who does this situation remind you of?”

2. “Based on what happened last time we tried this, what is your brain predicting will go wrong now?”

By bringing these implicit prototypes into conscious awareness (System 2), the leader can acknowledge the validity of the past prediction while explicitly differentiating the present context (Kahneman, 2011).

1. Current State Awareness (The Weather)

While “Geology” focuses on deep, stable memory structures, “Weather” focuses on the volatile, immediate physiological state of the team. Drawing on Feldman-Barrett's (2017) theory of allostasis, we recognise that “affect” (mood) is a biological barometer of the body's current energy budget. When a team is metabolically depleted—tired, stressed, or burnt out—their ability to process new information or update their predictions is compromised. They are more likely to rely on rigid, defensive heuristics (System 1) to save energy (Feldman-Barrett, 2017; Kahneman, 2011).

Traditional leadership often ignores this, pushing for high-cognitive-load change initiatives when the team's “weather” is stormy. This results in inevitable rejection, not because the idea is bad, but because the collective brain lacks the metabolic resources to encode it.

Leaders must move beyond tracking tasks to tracking metabolic solvency. If Practice 3 (below) is about stopping a leak, then this practice is about checking the hull's integrity before leaving port. It is a proactive resource conservation strategy to ensure the energy budget exists before a load is applied (Ohira, 2018; Shih and Yeh, 2024).

1. The Safety Prediction Index: Before high-stakes meetings, employ a simple index asking the team to predict their allostatic capacity: “Do we have the metabolic solvency (energy budget) to process this today, or are we running on empty?” (Feldman-Barrett, 2017; Sterling, 2012).

2. Pulse Checking: Recognise that silence often signals metabolic withdrawal (freeze response) (Ayub et al., 2021; Ohira, 2018). If the “weather” is poor, the leader must delay complex cognitive demands. Proceeding when the collective energy budget is insolvent guarantees prediction error and resistance (Hobfoll et al., 2018; Joo et al., 2023; Ohira, 2018).

3. Monitoring Prediction Error (The Diagnostic)

When a leader's action violates a follower's prediction, the brain experiences a “prediction error.” As Kahneman (2011) notes, surprise forces the brain out of efficient System 1 processing into effortful System 2 processing. In the Predictive Leadership model, this transition is a haemorrhaging of metabolic resources (Feldman-Barrett, 2017). Every moment the error remains unresolved, the follower's brain burns glucose trying to resolve the dissonance (Hoemann et al., 2019a; Lopez-Valeiras et al., 2022). Sarcasm, silence, or confusion are not just “bad attitudes”; they are the visible smoke indicating a fire in the predictive machinery and a rapid depletion of the team's energy budget. Intervening here is critical; leaving these signals unaddressed converts a temporary prediction error into the chronic “systemic drag” that drains future adaptive capacity.

In many organisations, leaders misinterpret these signals. Sarcasm, silence, eye-rolling, or sudden shifts in tone are often labelled as “resistance” or “bad attitude.” In the Predictive Leadership model, these are diagnostic data points. They signal that the leader's behaviour has clashed with the follower's internal model (Butko et al., 2025).

Leaders must be trained to catch these signals immediately and intervene, not to correct behaviour, but to investigate the prediction. The intervention shifts from “Why are you reacting that way?” to “I'm sensing a disconnect. What exactly did you expect to happen just now, and how did reality differ?”

This question validates the follower's internal model while forcing the error into conscious awareness, allowing the brain to update its prediction rather than entrenching the resistance. This shifts the leader's focus from judging the individual's reaction to auditing the organisational processes that caused the prediction error in the first place.

1. Scaffolding Future Predictions (The Architecture)

If the brain is a prediction machine, the leader's primary role is to be an architect of certainty. Uncertainty is metabolically taxing; it forces the brain to run multiple simulations to prepare for various outcomes. Leaders reduce this “tax” by providing clear concepts and reliable patterns that allow followers to predict efficiently (Iddrisu, 2025; Salameh-Ayanian et al., 2025). This practice draws on Barrett's concept of “Social Reality.” Leaders use language not just to describe the world, but to create the concepts that followers use to interpret it. By consistently linking actions to specific values and outcomes, leaders build a reliable scaffold that supports the team's future predictions.

However, leaders must provide the specific conceptual vocabulary required for followers to construct the future. As prior work has noted, leadership frameworks often overlook the role of emotional granularity (Butko et al., 2025); within the present model, this matters specifically because imprecise emotional labels increase prediction error, whereas clear naming helps stabilise future expectations. Hoemann et al. (2019b) argue that emotion words serve as invitations to form categories, allowing the brain to impose functional similarities on disparate sensory inputs. Without this linguistic scaffolding, the collective brain cannot resolve the prediction error caused by the new strategy, leading to sustained uncertainty.

To prevent future errors, leaders must “close the loop” on decisions. When a decision is made, the leader must explicitly state the “why” and the expected outcome, effectively programming the team's predictive model. Leaders cannot simply erase the “Geology” of past predictions. Instead, they must architect new predictions by “seeding” new concepts—using consistent language and visible closure to build a new, low-energy pathway for the brain to follow.

1. The Prediction Audit: Before complex decisions, leaders should document their own System 1 predictions (“I assume the team will love this”) and test them against the “Geology” and “Weather” of the team. This forces the brain to incur the tax of processing data over instinct. It is effectively a “Reality Check” that prevents the team from running on outdated software.

2. Public Closure: When an outcome deviates from what was expected, the leader must publicly acknowledge the error. “We predicted X, but Y happened. Here is why.” This prevents the team from forming superstitious or inaccurate associations (heuristics) that will distort future behaviour.

3. Standardise Risk Contexts: Use consistent protocols for high-uncertainty events (e.g. restructuring announcements) to minimise the “startle response.” Explicitly link decisions to established values to create a predictable environment where the signal of safety is continuous.

Just as the individual brain seeks to minimise metabolic cost, the organisation functions as a collective energy system (Feldman-Barrett, 2017; Sterling, 2012). Widespread prediction error (uncertainty) creates systemic “metabolic drag,” observable as change fatigue and cultural resistance (Ohira, 2018). Unresolved prediction error is a physiological stressor that contributes to psychosocial risk (Kjærgaard et al., 2024). Predictive Leadership is the operational response to stop this energy leak.

This model offers a novel approach to diversity and inclusion. Bias is often treated as a moral failing, but physiologically, it is a resource-conservation strategy (Feldman-Barrett, 2017; Russen et al., 2025). The brain uses a stereotype (a learned prediction) because it is a metabolically cheap System 1 shortcut. It costs almost nothing to run. Conversely, processing individual traits requires high-cost System 2 attention (Kahneman, 2011; Russen et al., 2025). In environments characterised by uncertainty or low information, the brain prioritises metabolically “cheap” surface-level cues over “expensive” deep-level processing. For example, Alon et al. (2023) found that in global virtual teams, leadership emergence was predicted by surface characteristics such as language proficiency and age, rather than deeper competencies like emotional intelligence, illustrating the brain's reliance on efficient, albeit sometimes inaccurate, prototypes.

Predictive Leadership treats bias as a stubborn prediction that persists because the metabolic cost of updating the model is high (Clark, 2013). Leaders must therefore “force” this expenditure using the Prediction Audit. Leaders explicitly document their intuitive prediction about a candidate and compare it against structured criteria. This forces the brain to incur the metabolic tax of processing data over instinct, effectively overwriting the cheap prototype with accurate, albeit expensive, data (Kahneman, 2011; Russen et al., 2025).

Change fails when it ignores the “metabolic tax” of uncertainty (Ohira, 2018). Predictive Leadership suggests that resistance is often a physiological (but unconscious) refusal to incur the energy cost of updating a deep-seated model. Leaders can mitigate this by validating the old prediction first (“I know your brain is predicting failure because of the 2019 restructure … ”) before introducing the new data. This lowers the threat response, making the brain more receptive to updating. By addressing prediction errors early, leaders can prevent the accumulation of the “systemic drag” described earlier—where repeated metabolic costs compound over time—thereby preserving collective cognitive bandwidth and reducing downstream resistance to change. Unresolved prediction error triggers specific cognitive processing styles that can stall organisational change. Isbell et al. (2016) demonstrate that negative, high-uncertainty emotions such as fear prompt a shift toward “local” or concrete processing. This suggests that the “metabolic tax” of resistance manifests cognitively as an inability to see the strategic big picture, forcing employees to fixate on minor details as a protective mechanism.

1. Manage prediction error, not attitudes: Leadership interventions are most effective when they focus on reducing uncertainty and mismatched expectations, rather than attempting to correct perceived motivation, mindset, or commitment. By minimizing the “metabolic tax” of surprise, leaders preserve the cognitive resources required for performance (Barrett, 2017; Kahneman, 2011).

2. Treat resistance as an energy signal: Apparent resistance or disengagement often reflects cumulative cognitive and metabolic load rather than defiance. Addressing unpredictability and overload restores capacity more reliably than increased pressure or monitoring, which only exacerbates the physiological stress response (Hobfoll et al., 2018; Ohira, 2018).

3. Use language to stabilise expectations: How decisions, priorities, and past events are named directly shapes future predictions. As Butko et al. (2025) argue, “emotional literacy”—the ability to precisely label emotional states (e.g. distinguishing “frustration” from “insecurity”)—allows for tailored responses that reduce the metabolic cost of ambiguity (Hoemann et al., 2019b).

4. Make recovery part of performance design: Sustainable performance depends on protecting recovery following periods of high demand. Leaders must recognize that chronic prediction error depletes “metabolic solvency,” necessitating periods of stability to restore the team's allostatic budget (Kjærgaard et al., 2024; Sterling, 2012).

5. Close the loop to update internal models: Explicitly explaining outcomes and decision rationales helps teams recalibrate expectations. This “prediction audit” updates the team's shared memory (“The Geology”), lowering the future cost of guessing and reducing compounding uncertainty (Gershman, 2017; Zacks et al., 2022).

Leadership is not just about influence; it is about stewardship of the collective brain. By understanding that the brain is a prediction machine designed for metabolic efficiency, leaders can move from reacting to downstream behaviour to architecting the upstream expectations that drive it.

We have argued that established models, while valuable, often fail to account for the “Geology” of past experience and the “Weather” of current physiological states. When left unaddressed, these repeated micro-costs accumulate into systemic drag, constraining not only individual recovery but the organisation's overall adaptive capacity by progressively narrowing the energy available for learning, coordination, and change. Predictive Leadership fills this gap. It provides a prescriptive framework—Mapping Past Experience, Awareness of Current State, Monitoring Prediction Error, and Scaffolding Future Predictions—that translates complex neuroscience into coachable behaviours.

Future research should operationalise and validate the framework, yet the approach already offers a practical way to navigate complexity, foster engagement, and improve performance by proactively managing predictions.

Ethical approval was not required for this study as no empirical data were collected and no human participants were involved. This is a conceptual paper proposing a theoretical framework based on existing literature.