Current node
Node
Epistemic Reliability vs. Moral Accountability
If a friend lies to you about the weather, you feel betrayed; if a thermometer is broken, you simply buy a new one. This distinction captures the heart of the debate surrounding Mona Simion’s **epistemic functionalism**. The central tension is whether knowledge-sharing is merely a mechanical transfer of information or a deeply human act of "giving one's word," requiring a moral agent who can be held responsible for the truth.
## The Functionalist Framework
In her work, such as [*Shifty Speech and Independent Epistemology*](https://academic.oup.com/book/36423), Simion argues that the "function" of communication is to generate knowledge in the hearer. To Simion, an agent—whether biological or artificial—is a good informant if they are **functionally reliable**.
Under this view, a machine does not need a "soul" or moral agency to participate in the epistemic community. It only needs to perform its function: providing accurate information. As Simion notes regarding the nature of epistemic functions:
> "The function of the testimonial device is to generate true belief in the hearer; as long as the device is reliable, the function is being performed."
> — Mona Simion, *Epistemic Stratification*
## The "Assurance" Critique
Critics argue that by focusing strictly on reliability, Simion overlooks the **Assurance View of Testimony**. Thinkers like [Richard Moran](https://philosophy.harvard.edu/people/richard-moran) and Edward Hinchman suggest that when we tell someone something, we aren't just acting as a "sensor"; we are offering a guarantee.
In human knowledge-sharing, the speaker invites the listener to trust them. If the information is false, the speaker incurs a **moral debt**. Because an AI cannot feel shame, face social repercussions, or "care" about its reputation, it cannot offer this guarantee. It cannot "stand behind" its words. For those who believe knowledge-sharing is a social contract, a machine’s output is mere "data," whereas a human’s output is "testimony."
## Does Functionalism Overlook Responsibility?
Simion’s likely defense is that "moral responsibility" is a separate category from "epistemic success." From a university lecturer’s perspective, we can break this down into two distinct roles:
1. **The Epistemic Role:** Does the source provide truth? (The AI excels here).
2. **The Moral Role:** Can the source be blamed for error? (The AI fails here).
Simion argues that for the purpose of *knowledge acquisition*, the moral role is often "epistemically redundant." If a highly advanced AI is 99.9% accurate, it provides more "knowledge" than a fallible human friend, even if the friend has the best moral intentions.
## The Emerging "Responsibility Gap"
The danger in Simion's approach is the [Responsibility Gap](https://en.wikipedia.org/wiki/Ethics_of_artificial_intelligence#Responsibility_gap). If we treat AI as a primary source of knowledge (functionalism) but cannot hold it accountable when it "hallucinates" or provides biased data, we risk a society where no one is responsible for the "truth" anymore.
Does the lack of a "punishable" agent degrade the quality of the knowledge itself? Or is our insistence on "moral responsibility" just a biological bias—a "human-centric" requirement that we should discard as we move toward a more integrated digital epistemology? These questions suggest that while Simion’s functionalism is efficient, it may struggle to explain the **indignation** we feel when a "reliable" system leads us into error.