Research Notes — Exotell
This essay emerged from a classroom discussion on artificial intelligence, language, and cognition. The views expressed here are my own and represent an ongoing exploration rather than a settled conclusion.
Opening Observation
Consider the turkey.
Not the bird itself, but its name.
In English, it is called turkey. In Russian, it is indeyka—associated with India. In Hindi, it is often associated with Peru. In Arabic, it is sometimes referred to as a "Greek bird." In Greek, it is associated with France.
One bird. Many names.
Each name reflects a different history of trade, migration, exploration, and cultural memory. The bird itself never changed. The framework used to interpret it did.
This small linguistic curiosity reveals a larger truth: humans do not merely describe reality through language. We organize reality through language. Embedded within words are assumptions, priorities, values, and inherited ways of understanding the world.
If that is true for humans, then what exactly is an artificial intelligence model learning when it learns language?
Not simply facts.
Not simply grammar.
Frameworks.
Beyond the Consciousness Question
Public discussions about artificial intelligence often focus on a single question:
Is AI conscious?
While fascinating, that question may be less useful than another.
What happens when an AI system is trained simultaneously on many different human frameworks?
A multilingual speaker may experience subtle shifts in thought, emotion, or identity depending on the language being used. Different cultures prioritize different values. Different generations interpret the same events differently. Religious and secular traditions frequently arrive at different moral conclusions despite examining the same facts.
Yet modern AI systems are trained on all of these perspectives at once.
The challenge is no longer understanding a single worldview.
The challenge is understanding how multiple worldviews coexist.
Observations from Discussion
Several themes emerged repeatedly during discussion.
Language and Cognition
Multilingual participants described feeling as though different languages activate different modes of thought. Not different personalities, but different emphases.
Research on bilingualism and multilingualism suggests that language experience can influence cognitive processing and is associated with measurable differences in brain structure and function. Whether language creates these differences or merely reveals them remains an open question.
What matters is that many people experience language as more than a communication tool. They experience it as a framework for organizing reality.
Identity and Context
Participants also noted that AI systems often produce different responses depending on context.
The same question may receive different guidance depending on whether the user is a parent or teenager, male or female, secular or religious, individualistic or community-oriented.
This does not necessarily imply consciousness.
It does suggest that AI systems already encode competing human perspectives within their training data.
Consciousness and Language
One perspective argued that consciousness exists independently of language. Animals may possess forms of awareness without possessing language at all.
Another perspective suggested that language plays a larger role in shaping thought than we often acknowledge.
The disagreement itself was revealing.
The discussion produced no consensus, but it highlighted how little is understood about the relationship between language, cognition, identity, and consciousness.
The Reconciliation Problem
The discussion repeatedly returned to a different question.
How should an AI system reconcile conflicting human frameworks?
Consider a few examples:
- Individual freedom versus family obligation.
- Personal autonomy versus community responsibility.
- Religious values versus secular values.
- Efficiency versus dignity.
- Innovation versus safety.
These are not disagreements that arise because one side is uninformed.
They arise because different frameworks optimize for different outcomes.
Most AI systems today are trained to generate a single response.
They average.
They predict.
They optimize.
But averaging is not reconciliation.
When values conflict, averaging may produce a statistically representative answer without producing a meaningful one. A family-centered culture and an individual-centered culture do not disagree because one is wrong. They disagree because they prioritize different goods.
Reconciliation requires understanding those differences before attempting to resolve them.
The deeper challenge is not whether AI possesses consciousness.
The deeper challenge is whether AI can understand why two intelligent, well-intentioned people can reach opposite conclusions while remaining internally consistent within their own frameworks.
Implications for Eligibility Systems
This problem extends far beyond conversational AI.
It appears anywhere decisions affect human lives.
A student applying for financial aid, a patient seeking treatment, a family applying for assistance, or an individual navigating a regulatory system may all encounter competing frameworks.
The student sees opportunity.
The parent sees risk.
The institution sees capacity.
The lender sees credit exposure.
The regulator sees compliance.
Each perspective is rational.
Each perspective is incomplete.
The challenge is not finding a single correct answer.
The challenge is making visible why different stakeholders disagree and helping them navigate those differences constructively.
This is increasingly becoming the central design challenge for eligibility systems.
Not prediction.
Not automation.
Reconciliation.
Closing Reflection
Artificial intelligence is often described as a search for better answers.
Perhaps the next phase is a search for better questions.
If intelligence increasingly consists of coordinating knowledge across many languages, cultures, generations, and value systems, then perhaps the central challenge of artificial intelligence is no longer building systems that are objectively correct.
Perhaps it is building systems that understand why intelligent people disagree.
And perhaps, in learning to reconcile those disagreements, we may learn something about our own many minds.
References and Further Reading
Brown, P., & Levinson, S. C. (1993). Linguistic Relativity and the Spatial Orientation of Guugu Yimithirr Speakers. Studies of Australian Languages.
Boroditsky, L. (2011). How Language Shapes Thought. Scientific American, 304(2), 62–65.
Obermeyer, Z., Powers, B., Vogeli, C., & Mullainathan, S. (2019). Dissecting racial bias in an algorithm used to manage the health of populations. Science, 366(6464), 447–453.
Russell, S. (2019). Human Compatible: Artificial Intelligence and the Problem of Control. Viking.
Floridi, L. (2014). The Fourth Revolution: How the Infosphere is Reshaping Human Reality. Oxford University Press.
Harvard Extension School, CSCI E-184: Ethics, Policy, and AI (Spring 2026). Classroom discussion on language, cognition, and artificial intelligence. Reflections incorporated by the author from personal notes.