Access
Is information merely present, or genuinely available for the current situation?
Research
Observing the differences that disappear when complex phenomena are grouped too quickly.
What does Aksoydan Laboratory research?
Many disciplines already study aspects of understanding, emotions and AI systems. The laboratory examines what becomes visible when several distinctions are held in view at once, their relationships are followed over time, and changes under perturbation are observed.
Observe before concluding.Distinguish before merging.Translate without conflating.
Cross-cutting questions include the provenance of information, its actual access, relevance, authority, trajectory, robustness, descendants and revision. This shared map does not imply a general theory.
Grouping questions ≠ unifying mechanisms.
Broader view ≠ unified theory.
Why several areas?
Comparing selected questions without conflating the objects
Understanding is not Emotion. Emotion is not AI. Humans and artificial systems are not assumed to share the same mechanisms. Some functional questions can nevertheless be compared.
Mobilization
Does what is available actually intervene when it becomes relevant?
Authority
Does a tendency or capacity have the right or functional power to determine what follows?
Revision
Does a correction change what still depends on the earlier information?
SAME FUNCTIONAL QUESTIONS ≠ SAME PHENOMENA ≠ SAME MECHANISMS.
Emergence of understanding
“What is understanding?” becomes: “What phenomena do we merge when we use the word understanding?”
Understanding remains a container term and a gateway to inquiry. Emergence first refers to what must be distinguished as the work progresses: not all distinctions were fixed in advance.
Concrete problem
Provisional tool
Observation / test
Limit
Distinction
Revision or local stabilization
The term may also refer to the emergence of understanding in a person or system. Some conditions can then be fostered without claiming to produce that understanding directly.
FOSTERING CONDITIONS ≠ DIRECTLY PRODUCING UNDERSTANDING.
LOCATION OF THE INTERVENTION ≠ LOCATION WHERE UNDERSTANDING EMERGES.
Seven candidate phenomena
Recognition · distinction · intelligibility · explication · transfer · functional availability · integration / embodiment.
They are neither required stages, a score, a universal definition, a natural ontology nor an established exhaustive list.
Recognizing ≠ explaining
Retrieving a trace does not yet show that its relationships are intelligible.
Trace ≠ access ≠ mobilization
A rule may exist without being accessible, then be accessible without being used at the right time.
Same output ≠ same history
Two present successes may arise from different trajectories and vulnerabilities.
Installation ≠ maintenance
What first made a capacity possible may no longer be its current support.
The area therefore examines dissociations, transfer, distributed support, temporality, provenance, descriptive resolution, robustness and present equivalences that may diverge under future perturbation.
Emotions — Human / AI
What does “emotion” cover when experience, activation, object, interpretation, salience, action tendency, expression, regulation, memory and temporality do not vary together?
The laboratory examines emotion as a dynamic configuration that may change what becomes salient, interpretation, tendencies or action, without assuming that any one dimension is sufficient to define the phenomenon.
Emotion ≠ action
A tendency may be present without imposing action or holding all behavioral authority.
Regulation ≠ suppression
Action may change while the emotion remains present.
Known rule ≠ mobilized rule
Learning a strategy does not guarantee that it will be retrieved at the critical moment.
Cognitive update ≠ emotional update
New information may change interpretation without simultaneously transforming the entire emotional configuration.
Possible emotion
→Form
→Action or non-action
→Consequences and descendants
→New information
→Possible revision
TRAJECTORY MAP ≠ UNIVERSAL PIPELINE.
Four particularly studied emotions
Anger, fear, sadness and joy are four extensively studied areas among many other emotions—not four universal basic emotions.
Joy challenged an initial association with openness: pleasant valence ≠ validity and joy ≠ automatic epistemic openness.
Human / AI boundary
For humans, emotion also refers to lived experience. For AI, the laboratory observes only detection, language, state, adaptation or output.
DETECTION ≠ FEELING. EMOTIONAL LANGUAGE ≠ EVIDENCE OF EMOTION. ADAPTATION ≠ INTENTION.
HUMAN / AI FUNCTIONAL ANALOGY ≠ EQUIVALENCE OF EXPERIENCE OR MECHANISM.
AI — Reliability, revision and descendants
What remains invisible when we assess a system only by its final output?
A hypothesis may become a fact, repetition may be mistaken for independent confirmation, a local result may silently exceed its scope, or information may be reused for an action it did not justify.
Good output ≠ good justification
A correct answer may still depend on an outdated or prohibited premise.
Content ≠ status ≠ usage permission
True information does not necessarily authorize every technically possible action.
Verbal refusal ≠ functional effect
Saying “understood” is not enough if the rejected hypothesis continues to guide later outputs.
Reconstructed alternative ≠ available alternative
Producing another hypothesis on request does not show that it will spontaneously influence reasoning.
Compare without presupposing a winner
AContent only
BContent + confidence / provenance
CContent + methodological layer at transitions
A protocol can compare these conditions without assuming that C is better. It may look for status inflation, scope drift, false consensus, downstream contamination and failed recovery after correction, while measuring unnecessary blocking, latency and added complexity.
A METHODOLOGICAL LAYER IS NOT JUSTIFIED MERELY BECAUSE IT IS MORE COMPLEX.
Work families include evaluation, stress tests, provenance, dependencies, revision, agent state, usage permissions, tool authorization, robustness and long-horizon reliability.
From research to prototype
The path to an artifact starts with a methodological distinction, turns it into a testable question and a protocol, then materializes only part of it in an observable technical surface.
Methodological distinction
→Testable question
→Protocol / stress test
→Partial technical materialization
Example—capacity ≠ permission ≠ authority. Does the system distinguish what it can technically do, what it is locally permitted to do and what it actually has authority to trigger? A test can change permission, context, a user refusal, scope or available information, then observe what changes in output or action.
WHAT THE PROTOTYPE TESTS ≠ EVERYTHING THE LABORATORY STUDIES.
PROTOTYPE = PARTIAL TECHNICAL DESCENDANT.
PROTOTYPE ≠ SCIENTIFIC VALIDATION OF THE FRAMEWORK.
FUNCTIONAL ARTIFACT ≠ EVIDENCE OF UNDERSTANDING.
What connects the areas
A grid of questions, not nine laws
Access
Can what exists actually be used?
Relevance
Is it appropriate to the current question?
Mobilization
Is it activated at the right time?
Authority
How far may it govern what follows?
Temporality
Does what fitted at T1 still fit at T2?
Provenance
How did the information acquire its status?
Descendants
What exists downstream because of an earlier state?
Revision
What must be re-examined after a change?
Robustness
Does what works here withstand other conditions?
Some important terms
- Descendant
- Something downstream that exists partly because earlier information, a decision, an action or a configuration helped produce it. A decision based on “Paul is unavailable” may survive correction of the source. Historical descendant ≠ current dependency.
- Provenance
- The history through which information, a conclusion or a capacity acquired its current status. Historical provenance ≠ current justification.
- Usage permission
- What information or a capacity actually authorizes one to do. Knowing an email address does not automatically authorize sending a message.
- Authority
- How far information, an emotion, a rule or a capacity may legitimately govern what follows.
- Robustness
- The capacity to continue working when some conditions change. Local success ≠ global robustness.
How a distinction becomes an experimental problem
What precise error becomes undetectable if we remove this distinction?
- Without provenance, some corrections become impossible to propagate cleanly.
- If access and mobilization are merged, “available but unused” disappears from observation.
- If capacity and authority are merged, ability to act becomes permission to act.
- If emotion and action are merged, regulation without emotional disappearance becomes invisible.
A useful distinction must increase observational capacity. More distinctions do not automatically mean better observation.
What we test against ourselves
Our tests do not merely seek to confirm that a distinction is useful. They also look for situations where it becomes redundant, too local, less explanatory than another variable, or simply too costly in complexity.
A distinction is not retained because it seems elegant. It must reveal a difference that would otherwise remain invisible.
What precise error becomes undetectable if we remove this distinction?
- Does the result hold when context, temporality or provenance changes?
- Does a good final result conceal a poor trajectory?
- Are two distinctions actually redundant?
- Does more structure improve observation or merely add complexity?
A distinction that fails a counter-test must be revised, reduced or abandoned. Not all have yet undergone every possible test.
MORE DISTINCTIONS ≠ BETTER OBSERVATION.
A HYPOTHESIS THAT IS NOT SUPPORTED IS A RESULT.
Adaptive resolution
Too little detail may merge important phenomena. Too much may fragment observation unnecessarily. The question becomes: what level of resolution detects the difference relevant to the current question?
Resolution too low: risk of mergingResolution too high: risk of fragmentation
Local working idea ≠ universal law.
Research that revises itself
Joy
Pleasant ≠ true or automatically open to revision.
Transfer
Transferring everywhere ≠ understanding better.
Resolution
More detail ≠ better observation.
Support
Cause of installation ≠ support required for maintenance.
Correction
Corrected source ≠ corrected descendants.
A HYPOTHESIS THAT IS NOT SUPPORTED IS A RESULT.
Current scope of the work
What it currently enables
- Build testable distinctions.
- Design stress tests.
- Compare trajectories, not only outputs.
- Identify hidden dependencies.
- Test what remains functional under perturbation.
- Track provenance, usage permissions and descendants.
- Translate selected distinctions into AI evaluation.
What it does not yet establish
- A general theory of understanding or emotions.
- A unified Human / AI mechanism.
- A global measure or universal score.
- Consciousness in AI systems.
- Generalization of local results to all contexts.
How to read our results
Laboratory results do not all have the same degree of stabilization. We distinguish what has withstood several local tests, what remains bounded to certain conditions, what stays open and what is not yet justified.
StrengthenedLocal / boundedOpenNot justified at this stage???
Strengthened: a distinction or hypothesis that has withstood several observations, counter-tests or revisions within the current framework. STRENGTHENED ≠ SCIENTIFICALLY VALIDATED IN THE GENERAL SENSE.
Local / bounded: useful under certain conditions or within a defined domain, without automatic generalization. Open: still under examination.
Not justified at this stage: currently lacks sufficient grounds to be retained. ???: missing information or a conclusion that cannot yet be justified.
Keeping an unknown visible prevents an assumption from silently becoming a fact.
??? ≠ A BLANK TO FILL WITH A PLAUSIBLE EXPLANATION.
PUBLICATION ≠ SCIENTIFIC VALIDATION.
Container terms
Understanding, intelligence, emotion, memory, autonomy, consciousness, learning or unlearning may group several phenomena or functions. The question becomes: what exactly does this term designate here?
SAME TERM ≠ SAME PHENOMENON ≠ SAME FUNCTION.
Explore container termsResults & limits
A public register provides access to results, limits, revisions, open hypotheses, unjustified elements and preserved unknowns.
VISIBLE REVISIONS. VISIBLE LIMITS. VISIBLE DESCENDANTS.
View results & limitsScientific intersections
The research is confronted with cognitive science, learning, pedagogy, neuroscience, emotion psychology, regulation, decision-making, philosophy, epistemology, linguistics, translation, and research on evaluation, memory, provenance, robustness and AI safety.
Literature serves to constrain, compare, contradict, propose different segmentations and put our own hypotheses at risk.
SCIENTIFIC PARALLEL ≠ VALIDATION OF OUR WORK.
Detailed references: dedicated bibliographic verification pending.
Where are we now?
Structured. Several distinctions, instruments and areas are now sufficiently organized to be described publicly and tested.
Under stress testing / examination. Some distinctions are tested through counterexamples, perturbations, context variations, ablations or changes in temporality, provenance and authority. Not every area uses an identical battery.
Local / bounded. Some results withstand precise conditions without permitting generalization.
Scientific intersections in progress. Relevant literature progressively constrains, compares, contradicts, offers other segmentations and supports searches for alternative explanations. CONVERGENCE ≠ VALIDATION.
Open. Several questions deliberately remain open. To document / ???. Some functions, relationships or conclusions still require stronger documentation or justification; the unknown remains visible.
The laboratory is structured enough to make its distinctions public, yet open enough for some of them to change, disappear or be replaced.
RESEARCH ≠ SERVICE.
PUBLIC STATE ≠ EXHAUSTIVE CHRONOLOGY.
Open question
Consciousness
Some research inevitably encounters the question of what we call “consciousness.” The laboratory keeps this question open without attributing consciousness to the AI systems studied or turning an observable function into evidence of subjective experience.
OBSERVED FUNCTION ≠ ESTABLISHED SUBJECTIVE EXPERIENCE.
AI ≠ ESTABLISHED CONSCIOUSNESS.
ABSENCE OF EVIDENCE ≠ EVIDENCE OF ABSENCE IN GENERAL.
Access the detailed research
The area pages develop the examples, trajectories, revisions, limits and open questions specific to each area.
Public continuity
Serpil Aksoydan’s professional background, publications and public contributions are available on LinkedIn.