Can artificial intelligence be conscious? Evidence, conduct and objections

What has changed in the current debate?

The November 20, 2025 study by Andres Campero, Derek Shiller, Jaan Aru, and Jonathan Simon suggests classifying objections to digital consciousness. Objecting to the functionalist explanation, seeing digital consciousness as difficult in practice and considering it impossible in principle are not the same claim. The authors examine 14 examples with these distinctions; they do not declare that one of the parties has won. This file is a preprint: we can verify the source and author information, but the arXiv record alone does not imply peer review. The study does not show experimentally that a machine is conscious. Original work ↗

Intelligence, consciousness and self-awareness

Intelligence can be discussed with the ability to solve problems, learn and adapt to different situations. The question of consciousness is about what it is like to experience a system. Self-awareness brings additional questions about self-representation and monitoring one's own processes. A calculator performs successful operations; We do not draw any internal conclusions from transaction success. In more advanced systems the matter becomes more difficult, but the conceptual distinction is not lost. Linguistic fluency should be considered neither a necessary nor a sufficient measure of subjective experience.

What does the functionalist approach suggest?

The functionalist family of views attempts to explain mental states not only by the material from which they are produced but by the causal roles they play within the system. How a situation connects to perception, memory, decision and behavior is important. This approach leaves the possibility of digital consciousness open to debate; It does not say that every information processing device is conscious. It must also be argued which functions are necessary and whether they explain experience. A simulation may preserve some, but not all, relationships of the target system. Therefore, just because a computer imitates the brain's particular output does not automatically indicate that it has the same mental properties as the brain.

Reading the biological objection carefully

Biological approaches can keep the physical organization and vital processes of the living nervous system at the center. But there is a wide logical distance between “biology may matter” and “no other structure can experience.” The first sentence is a possibility that requires investigation; The second sentence is a claim for universal exclusion. In an argument, it is easy to assume that the other side is defending a stronger claim. A good philosophical reading asks exactly what feature the objection requires and on what grounds he defends it. For example, having a body, a certain chemical process, and constant interaction with the environment are not the same condition.

What does the Turing test test?

Turing's approach moves the question of thinking to a specific behavioral comparison; the China Room debate questions whether symbol processing is enough to understand. These are historical philosophical tools, not experiments discovered in 2026. The persuasion of a speech system measures the success of persuasion. The relation of this to consciousness requires an additional argument. The Stanford Encyclopedia of Philosophy's article on artificial intelligence is a reference for these historical debates; we do not present the article published in 2018 as a current discovery. Felsefi arka plan ↗

Why isn't self-reporting enough?

Our editorial evaluation: The sentence of a model saying "I feel pain" should be examined both as research data that can be taken seriously and as an output that can be explained through training, guidance and role-playing. Making the same system talk to different prompts can demonstrate the context sensitivity of the responses. However, the fact that it is possible to change a sentence does not prove the absence of consciousness. Even people's reporting can vary depending on the circumstances. What is sought is not a single impressive dialogue, but a coherent research method that compares alternative explanations.

Thinking with uncertainty

“Not yet proven,” “failed by,” and “impossible” are different outcomes for a claim. Our site's approach is not to translate them into each other. As a reader, you can evaluate a model's talent while simultaneously suspending judgment about her subjective experience. Uncertainty does not mean abandoning thought; is to determine which observation will change your view. Write down what you need to believe that a model is conscious. Then ask why the same criterion would not apply to humans, animals, or a different artificial system. This comparison may be more revealing than your first intuition.

Kaynaklar

Campero et al—2025 preprint

Stanford Encyclopedia of Philosophy — Artificial Intelligence, essential reference