Motivation

• Doxastic theories assume that only what other beliefs one holds determines whether a belief is justifiable.

• Foundations theories grant a privileged role to a subclass of beliefs—the epistemologically basic beliefs—in this determination.

• It has been argued there is no way to defend epistemologically basic beliefs.

• The first alternative is to retain the doxastic assumption but reject basic beliefs, leading to a coherence theory.

• A coherence theory denies the existence of a privileged subclass of beliefs; all beliefs have the same fundamental epistemic status.

• Justifiability of a belief is determined by how it coheres with the entire set of one’s beliefs.

• Different coherence theories vary by their account of the coherence relation, but all agree that justification depends on a belief’s relation to all beliefs, not a privileged subset.

• One motivation for coherence theories is that since foundations theories are false, something like coherence theory must be true, given the doxastic assumption.

• A second motivation is the Neurath metaphor: we are like sailors rebuilding a ship at sea, using our existing doxastic system of beliefs, including beliefs about how to modify beliefs.

• We cannot abandon all beliefs to start fresh; procedural beliefs guide belief revision but are themselves revisable if they conflict too much with other beliefs.

• The validity of an argument only requires we not accept premises and deny conclusion simultaneously; what to reject depends on our relative confidence.

• Sometimes we might reject the validity of the argument itself, revising our beliefs about what counts as valid inference.

• The general strategy is to make the entire belief set internally consistent, using procedural beliefs to guide its revision, while allowing revision of procedural beliefs themselves.

• Although the Neurath metaphor motivates coherence theories, the philosophical credentials of coherence theories should be considered independently from it.

• Part of the Neurath picture is mistaken: the assumption that belief change is always guided by beliefs is false.

• For example, we reason by rules like modus ponens without necessarily having beliefs about these rules.

• Many people, including non-philosophers, have few or no beliefs about the logical rules they follow in reasoning.

• Complex reasoning, such as induction, often proceeds by rules that even philosophers cannot fully articulate.

• We do not guide inductive reasoning by beliefs about the rules of induction.

• Therefore, belief change is not always governed by beliefs about how to reason; many native reasoning processes operate via unarticulated rules without explicit beliefs about those rules.

• It might be inessential to the Neurath picture that we hold beliefs about how to change beliefs.

• A coherence theory could be defended by claiming belief change proceeds by reasoning from existing beliefs, weighing conflicts by relative confidence.

• However, this is an incomplete account: belief change also involves acquisition of new beliefs via perception, which is not inference from other beliefs.

• Perceptual beliefs enter the doxastic system without mediation by beliefs about how to acquire them.

• Both perception and inference often occur “under the surface,” unmediated by procedural beliefs.

• A correct account of belief change cannot rely exclusively on existing beliefs; it must also incorporate new perceptual beliefs.

• The accommodation of perception is a major problem for any doxastic theory.

• The doxastic assumption holds that justifiability of a belief depends exclusively on what one believes.

• Foundations theories try to accommodate this by supposing perception results in beliefs about sensory experience which are self-justifying, allowing reasoning from these basic beliefs to perceptual knowledge.

• The difficulty is that perception usually does not issue in beliefs about sensory experience, so perceptual knowledge cannot be based on reasoning from basic beliefs.

• For coherence theories, it is unclear how perceptual beliefs enter the doxastic system under the doxastic assumption.

• Although foundationalism is false, this does not clarify the true account of perceptual knowledge.

• The core idea in the Neurath picture remains: beliefs are innocent until proven guilty.

• We tend to become disengaged from the original reasons for our beliefs, remembering conclusions but not reasons.

• Without memory of reasons, discrediting those reasons cannot discredit the belief itself.

• This disengagement is a fact of human cognition due to limited information processing capacity; remembering reasons as well as conclusions would overwhelm memory.

• This supports the idea of a web of interrelated beliefs, where justification depends on a belief’s secure niche in the network.

• This networked structure aligns with what coherence theorists find appealing.

• However, this structure is also compatible with foundations theories because of the role of memory.

• In foundations theories, memory is analogous to sense perception; seeming to remember gives a defeasible reason to believe what is remembered.

• When beliefs become disengaged from reasons and are stored in memory, recalling them involves believing them based on seeming to remember.

• Such memory beliefs gain evidential status independent of original reasons, and are held unless defeated by stronger reasons.

• When memory beliefs conflict, we weigh the strengths of mnemonic reasons, essentially measuring our confidence in those beliefs.

• Thus, the correct part of the Neurath picture is accommodated by foundations theories, and does not argue in favor of coherence theories.

• However, foundations theories fail because, in remembering, we rarely have beliefs about seeming to remember—i.e., we lack epistemologically basic beliefs.

• The real disagreement between coherence and foundations theories lies in whether we have epistemologically basic beliefs.

• Since we lack appropriate epistemologically basic beliefs, all foundations theories are false.

• If the doxastic assumption holds, then some coherence theory must be true.

• Alternatively, one can reject the doxastic assumption, thereby rejecting both foundations and coherence theories.

• The downfall of foundations theories stems from their inability to handle perceptual input solely in terms of beliefs.

• This suggests the real problem is the doxastic assumption, which will eventually be rejected.

• Before rejecting the doxastic assumption, it is necessary to survey possible coherence theories to see if they can solve the problem.