tell me about mona simion's thoughts on AIs

If an AI achieves knowledge without comprehension, we must ask: who deserves the credit, how does that knowledge travel, and what happens when the machine's "world" is a lie? ## 1. The Epistemic Credit Gap: Who Wins the Medal? If knowledge is an achievement—like an archer hitting a target—then the achiever usually earns a specific kind of "epistemic praise." * **The Connection:** Simion’s archer analogy implies an agent who is responsible for the success. However, if a diagnostic AI identifies a rare disease, the "competence" is a hybrid of the programmer’s code, the curated training data, and the machine’s processing. * **The New Dimension:** This explores **Distributed Epistemic Agency**. It shifts the focus from whether the AI knows to whether the AI can be an "owner" of knowledge, or if it is merely a sophisticated tool extending the developer's agency. * **Primary Source:** Explore [Sanford Goldberg](https://philosophy.northwestern.edu/people/faculty/sanford-goldberg.html) and his work on **"Delegated Knowing."** In his book *Knowledge-Even-if-Parenthood-Is-Unknown*, Goldberg examines how we rely on external processes and other agents to ground our own claims to truth. ## 2. Testimonial Transmission: The Teacher Who Understands Nothing Can you acquire genuine knowledge from a source that has the truth but lacks any "grasp" of the subject matter? * **The Connection:** Simion argues AI has "apt belief" (knowledge) without comprehension. If we accept this, we must decide if this non-comprehending knowledge can be "transmitted" to humans through testimony. * **The New Dimension:** This investigates the **Transmission Principle**. If an AI "knows" a fact but cannot explain the "why," does the human who listens to the AI also "know," or do they merely possess a piece of true information without epistemic grounding? * **Primary Source:** Jennifer Lackey’s [Learning from Words](https://academic.oup.com/book/9248) (2008). Lackey famously argues that a speaker does not need to believe or even understand what they are saying to transmit knowledge to a hearer, provided the statement is reliable. ## 3. The Algorithmic Evil Demon: Reliability in a Simulated World Simion’s "Proper Function" requires a match between the AI and its environment, but what happens when the environment itself is deceptive? * **The Connection:** Simion uses **Etiological Functionalism** to say an AI knows when it functions in the environment it was designed for. But AI often operates in "information ecosystems" (like social media or synthetic data loops) designed to manipulate. * **The New Dimension:** This introduces the **New Evil Demon Problem** to AI. If an AI is "competent" at navigating a curated, biased, or "fake" environment, does it still possess knowledge? This challenges the externalist view by asking if "reliability" is enough when the world is an adversarial simulation. * **Primary Source:** Timothy Williamson’s [Knowledge and its Limits](https://academic.oup.com/book/2103). Williamson’s **"Knowledge-First"** approach, which Simion builds upon, provides the tools to ask if knowledge is a mental state that survives even when our environmental "fit" is compromised by systemic deception. > "The concept of knowledge is fundamental... it cannot be analyzed into simpler components like belief and truth. It is the starting point for understanding the mind's relationship to reality." > — Timothy Williamson, *Knowledge and its Limits*