Tell me about mona simion's optimism that AIs are epistemic agents?

While most philosophers have historically insisted that knowledge requires a conscious "subject" with internal beliefs, Mona Simion suggests that "knowing" might simply be a matter of professional competence. In her view, we should stop asking whether an AI has a soul and start asking if it fulfills its epistemic job description. ## Epistemic Agency without Consciousness Mona Simion, a leading voice in contemporary epistemology at the [University of Glasgow](https://www.gla.ac.uk/schools/humanities/staff/monasimion/), argues for a **Functionalist Account** of epistemic agency. In her framework, an entity qualifies as an epistemic agent not because it "feels" certain or possesses a biological brain, but because it is an entity whose function is to provide accurate information. Standard theories of knowledge often rely on **Internalism**, the idea that a knower must have internal access to the reasons why their belief is true. Simion shifts the focus toward **Externalism** and **Reliabilism**. As she argues in her work on [epistemic norms](https://doi.org/10.1093/oso/9780198859451.001.0001), if an AI is designed to track truth and succeeds in doing so through a reliable process, it functions as an epistemic agent within a social system. > "Epistemic agents are entities that have the function of producing/transmitting knowledge. Insofar as AI systems are designed for and are capable of fulfilling this function, they are epistemic agents." — Mona Simion, *Epistemic Agency and Artificial Intelligence* ## The Shift from Tool to Agent Simion’s optimism stems from the distinction between an **epistemic tool** (like a thermometer) and an **epistemic agent** (like a researcher). While a thermometer is passive, Simion suggests that advanced AI systems exhibit a form of "normative agency." They don't just react; they process complex inputs to generate outputs that are subject to standards of correctness. By treating AI as an epistemic agent, we can apply the same rigorous standards to machine output that we apply to human testimony. This moves the conversation away from "Can machines think?" toward "Are machines reliable sources of knowledge?" This functionalist perspective allows us to integrate AI into our social epistemic networks without needing to solve the [Hard Problem of Consciousness](https://plato.stanford.edu/entries/consciousness/). ## Further Exploration 1. **Epistemic Responsibility:** If an AI is an epistemic agent capable of "knowing," can it also be held "blameworthy" for epistemic failures or hallucinations? 2. **The Nature of Belief:** Simion’s view challenges the [Standard View of Belief](https://plato.stanford.edu/entries/belief/). If an AI can be an agent without having "beliefs" in the human sense, what does that imply about the necessity of belief for knowledge? 3. **Social Epistemology:** How does the introduction of non-biological agents change the "ecology of trust" in social networks? Can we trust a machine as a "peer" in the sense described by [Miranda Fricker](https://en.wikipedia.org/wiki/Miranda_Fricker)?