#epistemology

#epistemology

One of the strangest things about large language models is not what they get wrong, but what they assume to be correct. LLMs behave as if every question already has an answer. It's as if reality itself is always a kind of crossword puzzle. The clues may be hard, the grid may be vast and complex, but the solution is presumed to exist. Somewhere, just waiting to be filled in.

Nyaya Philosophy, one of the six classical schools of Indian philosophy ( Shad-Darshanas), is known for its systematic approach to logic and epistemology. Rooted in the idea that correct knowledge (pramā) is key to liberation (moksha), Nyaya provides a structured method of reasoning and critical thinking to uncover the truth. The sage Gautama (also known as Akṣapāda) was the founder of Nyaya philosophy. He contributed significantly to the development of logical reasoning in Indian intellectual traditions around the 2nd century BCE.

Science is neither a collection of facts nor merely a process, but rather the combination of both. All at once, science is simultaneously the full suite of knowledge that we gain from observing, measuring, and performing experiments that test the Universe, as well as the process through which we perform those investigations and refine our conclusions based on an ever-increasing set of data.

The other researcher suggested that the methodology we use should, to a large extent, be dictated by our epistemological philosophy. For example, are you a positivist, interpretivist, a hypothetico-deductivist, a post-positivist or some other stance appearing on the list of epistemological perspectives? I imagine many readers of this blog, like myself, will be surprised by this stance. Since day one of my research methods training, I've been taught that it's the research question that should dictate your methodology...

Consider what happens when you observe anything external to yourself. Light reflects off an object, travels through space, enters your eye, and triggers photoreceptor cells. These generate neural signals that journey through multiple processing stages in your visual cortex, integrate with other sensory information and memory, and finally produce the conscious experience of "seeing." That's extraordinary mediation. Multiple transformation layers where information gets filtered, compressed, interpreted, and reconstructed. By the time you "see" something, you're experiencing a highly processed representation, not the thing itself.

This way, Latour thought he could analyse the behaviour of scientists and verify how discussions, negotiations, and rivalries shape what becomes "knowledge." After his inquiries, Latour concluded that scientists apply an awful lot of personal biases and human behaviours to so-called factually correct scientific research. For Latour, "facts" gain authority through social processes, institutional validation, and consensus-building. Not just through "objective" discovery.

The traditional epistemological categories - truth, belief, inference, and justification - are ill-suited to describe ChatGPT, which does not possess beliefs or human grounding.