Fitch's knowability axioms are incompatible with quantum theory

How can we consistently model the knowledge of the natural world provided by physical theories? Philosophers frequently use epistemic logic to model reasoning and knowledge abstractly, and to formally study the ramifications of epistemic assumptions. One famous example is Fitch's paradox, which begins with minimal knowledge axioms and derives the counter-intuitive result that "every agent knows every true statement." Accounting for knowledge that arises from physical theories complicates matters further. For example, quantum mechanics allows observers to model other agents as quantum systems themselves, and to make predictions about measurements performed on each others' memories. Moreover, complex thought experiments in which agents' memories are modelled as quantum systems show that multi-agent reasoning chains can yield paradoxical results. Here, we bridge the gap between quantum paradoxes and foundational problems in epistemic logic, by relating the assumptions behind the recent Frauchiger-Renner quantum thought experiment and the axioms for knowledge used in Fitch's knowability paradox. Our results indicate that agents' knowledge of quantum systems must violate at least one of the following assumptions: it cannot be distributive over conjunction, have a kind of internal continuity, and yield a constructive interpretation all at once. Indeed, knowledge provided by quantum mechanics apparently contradicts traditional notions of how knowledge behaves; for instance, it may not be possible to universally assign objective truth values to claims about agent knowledge. We discuss the relations of this work to results in quantum contextuality and explore possible modifications to standard epistemic logic that could make it consistent with quantum theory.