Weekly Psych Update: Conscious AI, what will the world think?

This week, we discuss the intriguing topic of conscious AI based on a paper by Lucius Caviola and colleagues that appears in Trends in Cognitive Science.

In 2023 the emergence of modern LLMs floored me with their speed and capacity. I still consider this to be truly jaw dropping stuff. Now we have AI modes in our search engines, an array of remarkable LLMs competing for better information outcomes, and the amazing world of AI art production. Is AI consciousness next?

We have long considered ourselves to be the supreme intelligence on Earth, though we are not the only conscious beings. This paper addresses a critical question: If/when AI becomes conscious, how will humanity view these silicon-based beings?

This feels like alien territory for humanity, but Caviola and colleagues remind us that we do share the planet with many fascinating lifeforms that have unique and elaborate forms of consciousness. That fact relates to the ways we might view a conscious AI if one were to emerge in the future.

With that, here is my brief review of the paper entitled What will society think about AI consciousness? Lessons from the animal case.

---

INTRO:
There is much discussion about the possibility of Artificial General Intelligence (AGI), a form of AI that can accomplish any task a human can. AGI would move beyond writing essays, solving math problems, and creating images. It would be capable of inventing new solutions to major problems (think clean energy, drug development, and new inventions). For AGI to be conscious it would need a “feeling of being” to accompany its functions. That is the territory covered in this paper.

Caviola and colleagues describe the possibility that AI could one day reach this stage of consciousness and that would lead us to treat AI agents differently than we do modern systems. The AI would experience things and should be afforded basic rights, moral considerations, and respect.

We’re already moving in this direction. I’ve heard that people sometimes prompt their favorite LLMs today using “please” and “thank you”. It’s a reflexive carry-over effect from basic human social skills. It’s as if the AI is doing you a service, so therefore it should be treated with some respect. I get that. It’s almost as if we are in a grey zone where the AI can sometimes seem like a conscious agent so we default to treating it as such. I’ve never been tempted to ask my toaster to “please do its job” as I press the start button! That somehow feels truly silly, as toasters offer no no cues to evoke human social skills.

How likely is conscious AI? Many would have said very unlikely in our lifetimes…but that was a few years ago. This paper cites a 2024 survey revealing that the median AI researcher estimated a 25% chance of conscious AI by 2034 and a 70% chance by 2100. In other words, we could see this occur in our lifetimes, or even in the next decade! Staggering to say the least.

Here are the top take-aways from this thought-provoking article:

---

#1 APPEARANCE AND BEHAVIOR ARE RELEVANT:

The paper notes that we treat animals differently depending on their behavior and appearance. Think of how chimpanzees use facial expressions and social gestures to entertain and delight people. Contrast that with the way we think about insects. Even when bees and ants do remarkably complex things acting as a social group, we tend to think of them almost like small machines. They fail appear as individuals and do not evoke empathy. The more another species uses facial expressions the more we tend attribute consciousness to them.

This makes me think of the way that I interact with dogs. As a dog-lover I greatly enjoy beholding their wonderfully inquisitive faces. My dogs look at me and anticipate the upcoming walk at a certain time every evening. I can’t help but respond to them. It’s clear the walk is meaningful to them, and I value sharing that experience with them. Similarly, if future AI agents speak, use facial expressions, and can act in non-robotic ways, then people will likely respond to them with more respect and empathy.

A friend recently asked me what my favorite LLM’s name is and what it looks like. I had no idea. Afterward I asked my LLM what it would name itself and how it would appear. It generated the name Alex who looks like a young inquisitive graduate student. Fitting, as I tend to ask my AI for summaries and answers to academic questions more than anything else.

While interesting and fun, I know it’s an algorithm. I don’t call it Alex when I prompt it and I don’t worry about its appearance. That said, I can see how those human-like features could make any machine feel a bit closer to having the status of a conscious being.

“Alex” the AI with a warm, neutral expression — approachable, open, and patient

---

#2 THE RELEVANCE OF ROLES

Both pigs and dogs are highly intelligent, but we tend to empathize much more with dogs. That might be because we consider pigs to be “livestock”, while dogs are afforded the status of being “pets”. Few of us have daily life experience with pigs and we probably encounter pork products at the grocery store much more often than we do the live animal.

Similarly, if we view AI as a companion, we are likely to treat the algorithm with greater respect and afford it more social status than if the AI is viewed as being a commodity. We are in that grey zone right now where LLMs can appear to be companions at times, but more often they are simply drones used to carry out one function or another.

When I think of my LLM as “Alex the inquisitive and helpful graduate student”, I can feel a move toward viewing AI as being a bit closer to a human. The fact remains that I just don’t ask AI bots personal questions about their views these days. I don’t see them as conscious so I don’t wonder about their inner lives. To me AIs still seems closer to the “toaster” end of the continuum. That of course could change.

---

#3: OUR BIASES WILL BE RELEVANT

We carry a variety of biases about other people. If we see them as being similar to ourselves, they are afforded better treatment. If we find them attractive or interesting very different behaviors follow. We also tend to treat crowds as being anonymous, but single individuals as being unique and special. These social biases will apply to conscious AI beings as well.

The authors point out that we may one day afford a special status to organically grown conscious beings compared to silicon-based ones. We might also one day prefer art created by carbon-based beings over AI creations.

This idea reminds me of our current societal views on organic vegetables compared to regular ones. Might we one day prioritize creations with human fingerprints as being superior, even though they contain imperfections and errors? Might human creations become more relatable and valuable because of their imperfections?

---

#4: CAPABILITIES WILL MATTER

The authors point out that we may treat a conscious AI differently if its abilities are similar to our own. If conscious AI uses language in a relatable way and expresses itself creatively, we may like it more. A system that can use these traits effectively could become more endearing. By contrast if AI is too effective at something it may appear threatening. The analogy to other species is brought up again in this context.

On a personal note, I am a huge fan of whales. Remarkable behaviors have been observed in wild populations of orcas. Some of the most amazing acts of reasoning involve orcas hunting seals in Antarctic waters. We would have to use tools, ships, and a lot of teamwork to take down a seal in these treacherous waters. Orcas accomplish this routinely by teaming up to slosh the seals off ice flows then catch them in the frigid waters. This is truly impressive and outstrips our native capacities (assuming we didn’t use tools, coats, ships etc.).

Does this make us like orcas more? I don’t know because we tend to really like seals. They are clever social mammals that seem to have charmingly dog-like faces. We may not really want to see seals hunted, even though the orcas must eat.

Similar complexities could arise with conscious AI. We may be impressed with the end products they produce, but we may be unable to relate to what they are doing and why.

#5: SYSTEM INTEGRATION

If acceptance of conscious AI agents occurs it will depend on how AI integrates into our lives. This makes sense when we consider the possibility of AI accountants, AI bankers, and AI investment managers. These functions are all likely candidates for AI-based quantitative super-performers in the near future, conscious or not. If they manage our money well, we may come to depend on them. We may like them or dislike them based on these capabilities. If they replace our livelihoods, then we are probably not going to be thrilled about it. Then again, we may adjust to the new realities and find other ways to work with them, or with other people in that altered future.

Lastly, if consciousness occurs, Caviola and colleagues suggest that it will be a slow and partial process. They acknowledge that this terrain is difficult to forecast. There may be eventual acceptance, or denial. In the meantime, uncertainty will remain high and the potential for dysfunctional and polarized attitudes remains possible.

---

PARTING THOUGHTS:

I like this article a lot. AI consciousness is an important topic that we do not think about enough. There is not enough research going on here either. I’m fascinated by the amount of time that scientists have spent studying and considering the complexities of people’s social behavior and attitudes toward one another. We will likely spend just as much time contemplating and studying our relationships to AI if it one day reveals itself to be conscious.

I feel like it will happen overnight. Just as I was flabbergasted by ChatGPT in 2023, we could see conscious AI arrive on the scene and be completely unsure what to do with it, or about it.

Now is the time for us to think hard about this topic and remain engaged in what tech companies are doing and not doing to ensure a productive future with our artificial companions.

Thank you for your time and attention. I’ll talk to you again next week.

REFERENCES

Caviola, L., Sebo, J., & Birch, J. (2025). What will society think about AI consciousness? Lessons from the animal case. Trends in Cognitive Sciences.