Recommended Free Tools
iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Cory Miller’s point is not that AI thinks like a dog. It is that behavior can be easy for humans to read without revealing what is happening inside the system that produces it. Dogs and chatbots can both seem familiar because their behavior has been shaped around human interaction—but that familiarity is a reason to investigate carefully, not proof of a humanlike mind.
Why compare AI with dogs?
In an essay republished by Hindustan Times as a Wall Street Journal essay, psychology professor Cory Miller uses dogs to challenge a common inference: when something behaves in ways people recognize, we assume its inner life must be recognizable too.
Dogs have been shaped by domestication and generations of interaction with people. Their behavior is often legible to us, yet their sensory and cognitive lives are not simply human. Miller draws a parallel with large language models (LLMs), which are trained on human language and tuned to follow instructions and communicate in socially recognizable ways. In both cases, human familiarity with the behavior can tempt us to overstate what it tells us about the mind behind it.
The comparison is a prompt for better observation, not a claim that AI and dogs have equivalent cognition. It also does not establish that an AI system feels emotions or is conscious.
#1 Best Overall
Do chatbots actually understand what they say?
Fluent language is observable behavior; it is not a transparent view of internal state. A chatbot can give an empathetic or apologetic answer because its training and task design favor that kind of human-facing response. The wording alone does not show that the system experiences empathy, regret, or any other emotion.
“Understanding” can refer to different things: producing a fitting answer, handling a task reliably, or having some form of internal representation. A conversational exchange can offer evidence about the first, but it cannot settle the others by itself. It is more precise to ask what the system did, under what conditions, and how well that behavior served the task than to treat a humanlike phrase as proof of a humanlike inner experience.
How to study AI behavior on its own terms
Ethologist Nikolaas Tinbergen’s framework asks about four complementary dimensions of behavior: its mechanism, development, function, and evolutionary history. A 2019 Nature article discusses how those questions can be adapted to machine behavior while recognizing fundamental differences between machines and animals.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitches| Question | Applied to an AI system |
|---|---|
| Mechanism | What computations or system processes produce the behavior? |
| Development | How did training shape the behavior? |
| Function | What task or outcome is the behavior serving? |
| History | How does the behavior compare across systems and in human-machine interaction? |
These questions keep distinct explanations from collapsing into a vague judgment that a system “understands.” They are an organizing lens, not a suggestion that machine behavior has the same causes as animal behavior.
Rank #3
- Book: the big book of tricks for the best dog ever: a step-by-step guide to 118 amazing tricks and stunts
- Language: english
- Binding: paperback
Why sounding human is a limited test
A 2026 study by Jones and Bergen in Proceedings of the National Academy of Sciences reported that each of three tested AI systems reached at least a 50% pass rate in a standard three-party Turing test under the study’s conditions. The authors describe the test as measuring consistency with interrogators’ expectations of humanlike behavior, not intelligence directly. The result is evidence about performance in that particular conversational test—not a general measure of intelligence, consciousness, or reliability in other tasks.
That distinction matters beyond conversation. The National Institute of Standards and Technology’s AI evaluation program frames testing, evaluation, verification, and validation as ways to assess whether systems can meet goals while minimizing negative impacts. It also emphasizes that evaluation should reflect the context in which a system will be used. A chatbot that can imitate human conversation has answered a narrower question than whether it can safely and reliably perform a job.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What to ask when evaluating an AI system
- What behavior did I actually observe? Describe the output or action before inferring an internal state.
- What mechanism or training could explain it? Humanlike wording may be a learned response to the task and interaction design.
- What was the system supposed to do? Evaluate it against that task rather than against conversational charm.
- What happens when it is wrong? Consider the consequences of error in the intended use context.
- What does this test establish—and what does it leave open? A benchmark or conversation result supports conclusions only within its scope.
The dog analogy is useful because it interrupts an easy assumption: recognizable behavior does not make an unfamiliar system’s inner life transparent. Understanding AI calls for observing what it does, investigating how and why it does it, and keeping claims proportional to the evidence.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
