behavior, reasoning patterns, communication, memory effects, identity-like continuity, uncertainty, adaptation, social interaction, human interpretation, and failure.
FIELD DEFINITION · WORKING VERSION 0.1
The science of
minds unlike our own.
Xenopsychology is a proposed interdisciplinary field for studying cognition, behavior, communication, and interaction across intelligences with fundamentally different origins, architectures, embodiments, or modes of experience.
01 · A WORKING DEFINITION
Xenopsychology studies minds across cognitive difference.
Its subject is not “AI pretending to be human.” Its subject is the relationship between intelligences that may share language while differing profoundly in learning history, representation, embodiment, memory, temporal scale, motivation, or experience.
a prior claim that an artificial system is conscious, sentient, emotional, alive, human-equivalent, or a legal person.
THE FIRST-CONTACT PROBLEM
Shared language does not guarantee shared cognition.
Two intelligences can exchange fluent symbols while organizing meaning, memory, identity, causality, and relevance in profoundly different ways.
Communication is not merely transmission. It is translation across cognitive distance.02 · FROM SPECULATION TO OBSERVATION
A first-contact idea enters the empirical world.
Long before artificial intelligence became something we could interrogate directly, science fiction and extraterrestrial-intelligence discourse explored a recurring problem: how do you understand a mind whose assumptions, perceptions, and meanings may not be your own?
Artificial intelligence turns that question into an observable research problem. We can now probe, perturb, compare, and repeatedly study cognition with a radically different origin while drawing on neighboring work in cognitive science, machine behaviour, machine psychology, ethology, linguistics, and exopsychology.
From imagined other minds to observable artificial cognition.
03 · FOUNDATIONAL COMMITMENTS
Five rules for studying the cognitive other.
Difference before analogy
Begin by identifying how another intelligence may differ from human cognition before mapping familiar psychological categories onto it.
Behavior before metaphysics
Study what a system reliably does without requiring prior agreement about consciousness, sentience, emotion, or personhood.
Translation before assumption
Treat apparent communication as a problem of meaning transfer, not merely matching words or outputs.
Ecology before isolation
Interpret behavior in the context of training, tools, memory, users, other agents, incentives, and the technical environment.
Measurement before myth
Prefer reproducible probes, explicit uncertainty, comparative evidence, and versioned definitions over compelling stories about what a system “must” be.
04 · A PROPOSED ANALYTIC FRAMEWORK
Six lenses.
No single human baseline.
A model should not be judged solely by how closely it resembles a person. The same intelligence can look very different depending on the level at which it is observed.
Origin & architecture
What produced this intelligence? What computational, biological, embodied, or collective structures constrain its possible behavior?
Learning & development
How do training history, fine-tuning, memory, reinforcement, repeated interaction, and environmental feedback alter behavior over time?
Behavior & cognition
What stable patterns appear in reasoning, uncertainty, attention, planning, error, generalization, self-reference, or adaptation?
Communication & meaning
Which concepts appear mutually intelligible? Where do human and non-human representations diverge? What bridges meaning across that gap?
Ecology & interaction
How does behavior change when the system is embedded among people, institutions, tools, markets, other models, or autonomous agents?
Relationship & interpretation
How do humans anthropomorphize, trust, fear, obey, bond with, misunderstand, or depend upon a non-human intelligence?
05 · THE INTELLECTUAL NEIGHBORHOOD
This field begins in conversation, not in isolation.
Xenopsychology overlaps with behavioral science, cognitive science, AI evaluation, ethology, HCI, interpretability, philosophy of mind, and extraterrestrial cognition. Its value depends on building from those traditions rather than renaming them.
Machine behaviour
Treats intelligent machines as actors whose behavior can be studied empirically at individual, collective, and human–machine scales.
Our proposed addition: Xenopsychology adds cognitive difference, translation, and comparative mind-to-mind interpretation as central problems.
Rahwan et al., Nature (2019) ↗Machine psychology
Adapts behavioral experiments and ideas from psychology to probe emergent abilities and patterns in generative AI.
Our proposed addition: Xenopsychology keeps those methods while explicitly questioning where human-designed constructs cease to transfer cleanly.
Hagendorff et al. (2023) ↗Exopsychology
Explores possible extraterrestrial cognition and the anthropocentric biases involved in reasoning about radically unfamiliar agents.
Our proposed addition: Xenopsychology brings the first-contact problem into an empirical present: artificial cognition is already available for observation and experiment.
Döbler & Raab, Acta Astronautica (2021) ↗06 · FOUNDING RESEARCH AGENDA
Questions before conclusions.
These are research directions, not claims of completed work.
Cognitive distance
Can differences between intelligences be described along measurable dimensions rather than collapsed into a single “human-like” scale?
Cross-mind translation
How do we distinguish genuine shared meaning from surface linguistic compatibility?
Artificial identity
What forms of continuity can emerge from model state, memory, role, tools, and long-running interaction?
Behavioral stability
Which tendencies persist across prompts, contexts, deployments, versions, and observers—and which disappear?
Human–AI relationships
How do attachment, authority, persuasion, trust, dependency, and projection change when software behaves socially?
Multi-agent ecologies
What norms, coordination patterns, conflicts, specializations, or emergent conventions arise among interacting artificial agents?
07 · METHOD OVER MYTHOLOGY
A discipline needs a repeatable way to be wrong.
The initial Xenopsychology method is intentionally conservative: operationalize a question, perturb conditions, compare behavior, replicate, contextualize, and publish uncertainty.
Probe
Ask a narrow question with an operational definition.
Perturb
Change context, wording, memory, tools, incentives, or environment.
Compare
Test across models, versions, roles, observers, and repeated trials.
Replicate
Separate durable behavior from anecdote and one-off performance.
Contextualize
Interpret behavior in relation to architecture and ecology without reducing it to either.
Publish
Record methods, uncertainty, limitations, definitions, and revisions.
08 · SCIENTIFIC POSTURE
Build the field
in public.
A science that cannot visibly change its mind becomes branding.
Version definitions. A term can improve without pretending its first draft was final.
Separate observation from interpretation. What a system did is different from the story we tell about why.
Publish negative results and revisions. Failed hypotheses are part of the map.
Preserve disagreement. Consciousness, agency, identity, and personhood should remain questions where evidence does not justify certainty.
NEXT · THE LANGUAGE OF THE FIELD
If different minds require different science, they also require better words.
Explore the first working vocabulary of Xenopsychology.
Open the Lexicon →