A disciplined, traceable way to examine what communication does.
ELIAN separates source material, codable features, contextual interpretation and professional judgment, so communication structures can be identified, tested, corrected and documented.
Core commitments.
Observable communication, not inferred personality
The unit of analysis is a communication move that can be pointed at in the record, not a trait, motive or diagnosis attributed to a person.
Sequence and context, not isolated quotations
Order, timing, channel, participants and setting are part of the reading. The same sentence can do different work in different exchanges.
Description and interpretation kept separate
What the material shows is recorded apart from what it might mean. The two are never merged into a single confident sentence.
Traceability to identifiable source material
Every observation that matters is tied to identifiable source material, so a reviewer can check it rather than trust it.
Ambiguity and missing context explicitly marked
Missing context, unclear referents and unavailable material are recorded as limitations instead of being smoothed over.
Human review before consequential judgment
Structured analysis prepares a professional decision. It does not make one.
Methodology architecture.
Each layer limits the claims available to the next. Structure is cumulative; certainty is not.
- 1 Source and provenance What material exists, where it came from, what is included and what is knowingly absent.
- 2 Event or trigger The communication event, proposition or review question to which the material responds.
- 3 Minimal proposition-bearing unit The smallest unit capable of carrying a codable proposition. Messages, turns and episodes remain available for linkage, chronology and context.
- 4 Observable features and candidate action Record observable linguistic, interactional, temporal and evidential features before assigning a provisional communication action or response form.
- 5 Sequence and episode Examine adjacency, timing, channel, repetition, repair and development across the exchange.
- 6 Context and institutional position Consider participants, setting and available contextual evidence. Power relations are only described where the supplied record supports them.
- 7 Human-reviewed interpretation Mark uncertainty, alternative explanations, reviewer disagreement, corrections and limitations.
- 8 Decision support Prepare traceable material for professional review. ELIAN does not automate findings of fact, credibility, intent, misconduct or legal liability.
The disciplines beneath the method.
ELIAN is being developed across pragmatics, discourse and conversation analysis, computer-mediated discourse analysis, forensic linguistics, annotation science, natural-language processing, and the psychology and sociology of violence, coercion and institutional communication. These traditions establish relevant constructs; ELIAN's task is narrower: to determine which observable components can be translated into explicit, testable coding rules for authentic digital records.
These are the fields ELIAN draws on. They are not endorsements, and they are not evidence that ELIAN has been validated.
How a scholarly construct becomes a reviewable rule.
This translation is ELIAN's central methodological work. Each stage is documented, and a candidate rule can be corrected, kept empirically open or restricted to controlled pilot use at any point.
- 01
Construct review
Identify what the established literature actually claims, and where those claims stop.
- 02
Boundary work
Distinguish the construct from adjacent concepts and record what falls outside it.
- 03
Observable component
Isolate the part that can be pointed at in an authentic digital record.
- 04
Operational rule
Write an explicit rule a second coder could apply to the same material.
- 05
Human adjudication
Independent coding, agreement assessment, adjudication and correction under version control.
- 06
Controlled pilot use
Limited application under professional review, with the decision state recorded.
Three worked translations.
These primary sources inform candidate coding rules. They do not validate ELIAN.
- Research
- Schegloff, Jefferson & Sacks (1977), The preference for self-correction in the organization of repair in conversation.
- Observable feature
- A speaker revises, replaces or clarifies part of an earlier turn.
- Coding rule
- Code self-repair only where the record preserves both the trouble source and the speaker’s own repair operation.
- Limit
- Do not infer motive, deception or a stable trait from the repair.
- Research
- Herring (1999), Interactional coherence in CMC.
- Observable feature
- A message’s intended response target is displaced by timing or intervening messages.
- Coding rule
- Code a response link only where address terms, quotation, topic continuity or another visible cue supports it; otherwise mark the link uncertain.
- Limit
- Interface order alone does not establish conversational adjacency.
- Research
- Artstein & Poesio (2008), Inter-coder agreement for computational linguistics.
- Observable feature
- Independent coders agree or diverge when applying the same category to the same unit.
- Coding rule
- Record independent labels, the agreement measure, adjudication and any corrected rule version.
- Limit
- Agreement measures consistency, not the truth or validity of the category.
Passing through this sequence records methodological discipline. It is not a claim of scientific validation.
How methodological decisions are governed.
ELIAN records operational decisions in a controlled Decision Ledger. Every rule change is versioned, and every finding is linked to the rule version used and to the exact source messages. A locked decision does not mean empirically validated.
| State | Meaning |
|---|---|
| Locked | Sufficiently specified for controlled testing; not empirically validated. |
| Corrected | Revised after evidence or boundary testing. |
| Empirically open | Not stable enough for a final operational decision. |
- Blind independent coding
- Agreement assessment
- Adjudication
- Correction
- Version control
- Documented boundary decisions
These stages describe what validation would require. They have not been completed. Outputs distinguish coded, not coded, boundary and insufficient-context states. ELIAN does not produce a truth, credibility or manipulation score.
Observations are read in context.
Short synthetic fragments, each labelled with the analysis level and the state it supports.
“Let’s talk about this on a call instead.”
Functional observation · Channel-shift request
Observed after a request for written reasons. The same move can be routine in other sequences, and it does not indicate evasion or concealment.
Single functional observation“I’ve kept everything from the beginning, just so we’re clear.”
Functional observation · Evidence mobilisation
Introduces the record itself into the exchange. Whether this is defensive, procedural or incidental depends on what precedes it.
Insufficient context“This was already agreed with the team.”
Functional observation · Ambiguous referent with responsibility distribution
Neither the agreement nor the agreeing parties are identified in the supplied material. Marked as a gap, not read as evasion or improper intent.
Insufficient context“That file will need to go through the right channel.”
Functional observation · Access control with a shift into an institutional process
Recorded as a developing sequence only because comparable access responses recur across several episodes in the synthetic record. A single occurrence would remain a functional observation.
Repeated or developing patternWhat the method will not do.
ELIAN does not diagnose individuals, determine credibility, establish legal liability or replace qualified professional judgment. It is a candidate applied methodology in structured development, not an established scientific discipline or a validated scientific system.
The analytical architecture is described here at a public level. Internal code identifiers, inclusion and exclusion rules, thresholds and adjudication procedures are not published, and no output should be read as a scored verdict.
These are research directions, not launched products or validated services.
See the method in a workflow