# Contested concept: evidence of consciousness

_**What you are reading**: one contested-concept page in Deliberus's deliberation graph. Deliberus maps human reasoning — including claims that are contested, mistaken, or extreme — so that disagreement can be examined precisely. A claim's presence here is not an endorsement: the graph records that someone argued it, how it connects to other reasoning, and how far scrutiny has gotten. The metadata on this page describes the state of that scrutiny, not the truth of the claim._

---

- **Contestedness**: criterion
- **Diagnosis**: Disagreement about evidential standards underlies the dispute over whether current LLMs are conscious.
- **Canonical**: `/concept/evidence of consciousness` (data: `/concept/evidence of consciousness.json`)

## Senses in use
- **General abilities and behavior count as evidence, and benchmarks might be found** — used in 5 claim(s)
- **Fragile or trained-in reports and human attribution tendencies are weak evidence** — used in 4 claim(s)
- **Verbal reports are a guide, in humans and AI** — used in 4 claim(s)
- **Compelling descriptions count only if the system was not trained on such material** — used in 3 claim(s)

## Claims using this term
- [What really matters [for evidence of consciousness in language models] is the system's behavior that prompts [the reaction that some langua…](/claim/claim_0ead70661e7e.md) _(sense: General abilities and behavior count as evidence, and benchmarks might be found)_
- [In humans, we rely on verbal reports. We use what other people say as a guide to their consciousness.](/claim/claim_25ae657e5619.md) _(sense: Verbal reports are a guide, in humans and AI)_
- [The fact that we are seeing increasing generality in large language models suggests a move in the direction of consciousness.](/claim/claim_2c8e5e3a2766.md) _(sense: General abilities and behavior count as evidence, and benchmarks might be found)_
- [When reports of consciousness [by language models] are as fragile [as language models reporting they are sentient and, with a single-word a…](/claim/claim_2dd3b97b1b72.md) _(sense: Fragile or trained-in reports and human attribution tendencies are weak evidence)_
- [As Schneider and Turner formulate the test, it's very important that systems not actually be trained on these features [of consciousness].](/claim/claim_5f3342ffe236.md) _(sense: Compelling descriptions count only if the system was not trained on such material)_
- [We should rely on verbal reports as a guide to consciousness in AI systems as well as in humans.](/claim/claim_67179ffc5f8f.md) _(sense: Verbal reports are a guide, in humans and AI)_
- [If [an AI system] has been trained on [material describing features of consciousness], the evidence [that the system is conscious] is much …](/claim/claim_69607ae81c52.md) _(sense: Compelling descriptions count only if the system was not trained on such material)_
- [If you get an AI system that describes features of consciousness in a compelling way, that's some evidence [that the AI system is conscious…](/claim/claim_8f84cf099eb8.md) _(sense: Compelling descriptions count only if the system was not trained on such material)_
- [We rely on verbal reports as a guide to consciousness in humans.](/claim/claim_b184f093c665.md) _(sense: Verbal reports are a guide, in humans and AI)_
- [The general abilities of large language models provide at least some initial reason to take the hypothesis [that large language models are …](/claim/claim_c6c4873c7300.md) _(sense: General abilities and behavior count as evidence, and benchmarks might be found)_
- [[The fact that some language models seem sentient to some people] is not strong evidence [that large language models are conscious].](/claim/claim_c9dfba0eb6d1.md) _(sense: Fragile or trained-in reports and human attribution tendencies are weak evidence)_
- [LaMDA said, in response to Lemoine's question whether it would like more people at Google to know that it is sentient: "Absolutely. I want …](/claim/claim_dce87129db80.md) _(sense: Verbal reports are a guide, in humans and AI)_
- [The fact that [LaMDA] has learned to imitate those claims [about consciousness made by people in its training corpus] doesn't carry a whole…](/claim/claim_eee2e1f26f9e.md) _(sense: Fragile or trained-in reports and human attribution tendencies are weak evidence)_
- [Can we find benchmarks for consciousness? That is, can we find objective tests that could serve as indicators of consciousness in AI system…](/claim/claim_f129c81987e1.md) _(sense: General abilities and behavior count as evidence, and benchmarks might be found)_
- [As many people have observed, two decades ago, if we'd seen a system behaving as LLMs do without knowing how it worked, we'd have taken thi…](/claim/claim_f4c98b98a597.md) _(sense: General abilities and behavior count as evidence, and benchmarks might be found)_
- [We know from developmental and social psychology, that people often attribute consciousness where it's not present.](/claim/claim_fbe4eefb2aa7.md) _(sense: Fragile or trained-in reports and human attribution tendencies are weak evidence)_