Epistemic Akrasia

Definition and Basic Phenomenon

Epistemic akrasia is the alleged phenomenon of being epistemically weak‑willed: believing a proposition while at the same time judging that, given one’s total evidence, one ought not to believe it (or ought to suspend judgment, or believe its negation). Formally, an agent is epistemically akratic when both of the following are true:

1. They believe a proposition p; and
2. They believe that, given their evidence, they should not believe p (for example, they should reject p or suspend judgment about p).

The concept is modeled on practical akrasia (weakness of will in action), where someone acts against their own all‑things‑considered judgment of what they ought to do. Epistemic akrasia transposes this tension into the realm of belief and justification.

A closely related phenomenon is higher‑order evidence conflict, in which an agent’s first‑order evidence supports p, but their higher‑order evidence (evidence about how they formed their belief, about their reliability, or about others’ disagreement) appears to support the claim that believing p is irrational.

Arguments for and against Possibility

Philosophers disagree about whether epistemic akrasia is merely psychologically possible but irrational, or even conceptually incoherent.

The Incoherence Objection

Many epistemologists argue that epistemic akrasia is epistemically impermissible or impossible for a rational agent. Their central claims include:

• Belief aims at truth and evidence-responsiveness.
  If a person believes p while sincerely judging that their evidence makes p unjustified, they are violating the norm that belief should track one’s assessment of the evidence.

• No stable rational perspective.
  On this view, it is not possible to occupy a single, coherent epistemic perspective in which “p” and “I should not believe p on my evidence” are both held as rational beliefs. One or the other must be mistaken.

• Transparency of belief.
  Some philosophers argue that when we ask ourselves “Do I believe p?”, we answer by consulting whether we take p to be supported by our evidence. Because of this “transparency,” it is claimed that genuinely judging that the evidence does not support p should, by rational necessity, undermine belief in p.

Proponents of this line frequently conclude that any apparent cases of epistemic akrasia are either unstable psychological states, self-deception, or misdescriptions of what the agent really believes or judges.

Defending the Possibility of Epistemic Akrasia

Others contend that epistemic akrasia is not only psychologically possible but may, in some special circumstances, be rationally permissible.

They offer several types of cases:

• Mismatch between levels of reflection.
  An agent might, at a first-order level, feel compelled by vivid evidence for p (e.g., perceptual or testimonial evidence), yet at a reflective, higher-order level conclude “Given how biased or tired I am, my evidence is probably misleading; I should not believe p.” This pull between levels may support akratic states.

• Higher‑order evidence and self‑distrust.
  When agents encounter strong higher‑order evidence that their reasoning is unreliable (e.g., discovering they have taken a drug that induces irrational confidence), they may rationally judge they should not believe p even while finding it psychologically irresistible to continue believing p. Some authors argue that, during this transition, a rational agent can briefly and perhaps even blamelessly occupy an akratic state.

• Disagreement cases.
  In deep peer disagreement, an agent might think:

  • “Given my assessment, p is true.”
  • “But given that an epistemic peer disagrees, my total evidence says I should suspend judgment.”
    If both beliefs are held concurrently, this yields epistemic akrasia.

Defenders of possibility often emphasize that rationality constraints may be weaker or more context-sensitive than the critics assume. They also distinguish between ideal rationality (which might forbid epistemic akrasia) and non‑ideal, human rationality, which may allow it as an understandable response to complex epistemic situations.

Connections to Rationality and Disagreement

Epistemic Rationality and Higher-Order Evidence

Debates about epistemic akrasia relate closely to theories of epistemic rationality:

• Uni-level views claim that all rational requirements on belief can be captured at a single level of evidential assessment. On such views, systematic conflict between first‑order and higher‑order judgments often appears irrational; thus, epistemic akrasia tends to be ruled out or heavily disfavored.

• Multi-level or hierarchical views allow that what one rationally ought to believe at a higher level (about one’s own reliability, methods, or evidence) may occasionally come apart from one’s first‑order doxastic states. This opens space for rationally excusable or even required akratic episodes, especially when higher‑order evidence is strong.

A key question is whether higher‑order evidence should override first‑order evidence. Those who think higher‑order evidence can be decisively weighty sometimes regard akrasia in transition as practically unavoidable: as one updates on higher‑order evidence, there can be moments where belief and judgment misalign.

Peer Disagreement and Steadfast vs. Conciliatory Views

Epistemic akrasia also figures in discussions of peer disagreement—cases in which two agents who are (roughly) equal in intelligence, evidence, and reasoning reach opposed conclusions.

• Conciliationists argue that discovering peer disagreement usually requires reducing confidence in one’s belief or suspending judgment. Some conciliationists claim that to persist in full belief while judging that one’s total evidence now favors suspension is a paradigm case of irrational epistemic akrasia.

• Steadfast views allow that one can reasonably retain one’s original belief despite peer disagreement. However, even steadfast theorists typically hold that one’s higher‑order judgments should align with one’s retained belief. Thus, a fully rational steadfast agent should not both keep believing p and also judge that, all things considered, they should not believe p.

Normativity and the Aim of Belief

Finally, the topic intersects with questions about the norms that govern belief:

• If belief is normatively constrained by an “ought” of evidence-responsiveness, epistemic akrasia looks like a direct violation of that ought.
• If, instead, epistemic norms allow for local inconsistency during processes of reflection, revision, or self-correction, then short‑lived or constrained forms of epistemic akrasia may be seen as part of ordinary, non-ideal rational cognition.

Philosophers remain divided over whether fully rational agents must completely avoid epistemic akrasia, or whether it is an unavoidable (and sometimes even appropriate) feature of human epistemic life, especially in the presence of complex higher‑order evidence and intractable disagreement.