Analysis Modes
Eight Ways to Read a Debate
A Metamorfon session produces raw material: several AI models respond to one another, advance positions, converge on some points and disagree on others. That material is rich, but dense. Without a reading framework, its complexity is difficult to grasp: it needs to be organised, its layers distinguished, and its salient phenomena brought into relief.
Analysis modes serve this function. Each applies a distinct perspective to the material of a session and produces a different kind of knowledge. They are not competing: they do not interrogate the session from the same angle, but they can complement one another. A single session can therefore be analysed successively through several modes, whose outputs illuminate different dimensions of the debate.
Analysis modes do not necessarily work from an entire session in raw form. They always receive the ten most recent turns in full verbatim form. Earlier turns, when they need to remain available as context, are provided in summarised form.
The user therefore chooses when to launch an analysis and, by doing so, which recent portion of the debate they want to audit directly. This distinction matters particularly in long sessions: analysis modes can take earlier context into account, but their most precise observations — the attribution of a concession, the emergence of a concept, the persistence of a tension, a shift in position — concern the ten most recent raw turns actually submitted for analysis.
Their conclusions should therefore be read at the scale of that corpus: a tension described as “persistent,” an identified emergence, or an attributed concession is established first within the raw analysis window, unless the earlier context supports a broader conclusion.
Eight modes are available. Here is the function of each.
Meta-Analysis
Meta-Analysis identifies the implicit structures at work in the debate: axioms accepted without justification, epistemic styles, shared biases and blind spots, and convergences or divergences in framing. It analyses what organises the positions within the frame of the session, rather than the manifest content of the arguments.
It is particularly useful when an agreement or disagreement appears self-evident: it shows the shared or divergent presuppositions on which it rests. It neither validates nor invalidates the positions; it situates their internal conditions. It is suited to political, philosophical, or strategic questions where framing matters as much as explicit arguments.
Integrative Synthesis
Integrative Synthesis reconstructs the progression of the session in a coherent form: initial framing, circulation of arguments, pivots, shifts, concessions, and points of arrival. It does not merely summarise each model’s responses separately; it restores the trajectory of the debate.
It produces a coherent account that organises the contributions without conflating or averaging them. This mode is particularly useful when the progression of a session needs to be preserved, shared, or reused in a readable form.
Emergence Analysis
Emergence Analysis identifies what appeared during the debate without having been present in that form at the outset: concepts forged during the session, unexpected distinctions, and conceptual articulations produced or transformed through friction between positions. This mode makes the productive dimension of a debate visible — not because novelty would be impossible for a single model, but because it allows its interactional construction to be traced.
It identifies these moments of emergence, names them, and makes their genesis explicit: how a formulation was sharpened under objection, how a distinction was taken up and then displaced, how several contributions eventually stabilised a shared object. It thus distinguishes what was actually constructed through the exchange from what was already present in the initial positions.
This mode is suited to conceptual research, philosophical exploration, and discussions intended to bring out distinctions or concepts absent from the starting point.
Tension Mapping
Tension Mapping identifies disagreements that persist throughout the session despite confrontation. It does not treat these disagreements as failures of the debate; it treats them as phenomena worth examining in their own right, revealing where positions continue to resist one another — whether the disagreement concerns facts, methods, values, or the very criteria by which the issue should be judged.
This mode finds a conceptual reference point in Jean-François Lyotard’s analysis of the differend: some tensions no longer concern only the answer, but the rule of judgement itself. Not every persistent divergence is a differend in the strict sense. Tension Mapping makes explicit the nature, depth, and resolvability of these disagreements, in order to understand what might still shift them — or what first requires the frame itself to be reconsidered.
It is suited to ethical, political, or aesthetic questions where plurality is inherent to the object and where seeking consensus at any cost would impoverish the debate.
Critical Archaeology
Critical Archaeology traces the conditions that delimited the debate before the models even began to argue: the wording of the question, the vocabulary used, presuppositions, exclusions, and framing choices. It also examines user interventions when they open, close, or redirect particular avenues in the discussion. Its purpose is not to judge whether the question was well posed, but to make visible the choices it already contained.
This mode therefore produces a defamiliarising effect. Categories that appeared self-evident emerge as the products of one particular way of carving up the problem, among other possible ones. It is suited to questions where the terms of the debate are themselves at stake, where framing appears to shape the responses, or where one wants to understand what had to be assumed for the discussion to take this form.
Horizon of Possibilities
Horizon of Possibilities identifies what the debate made formulable without actually formulating it. It works from traces left in the session — unresolved tensions, implications not yet drawn, nascent concepts, shifts in framing — to construct extensions that now have argumentative grounding.
It may thus bring out a distinction that several passages prepare without naming, a consequence that remained implicit, or a question that became more relevant by the end of the exchange than it was at the outset.
This mode is particularly useful when a session forms part of a longer investigation. It does not freely extend the debate: it formulates directions that the debate’s own trajectory has made defensible.
Argumentative Evaluation
Argumentative Evaluation judges how well the arguments were conducted during the session — the soundness of inferences, the handling of objections, the internal coherence of positions, and the ability to recognise one’s own contestable presuppositions. Together with Source Verification, it is one of two normative modes: the other six describe, while these two evaluate according to distinct criteria.
This evaluation concerns the manner in which positions are sustained, not the correctness of the conclusions or their factual verification. An argument may be well conducted in defence of a contestable thesis; another may be poorly conducted in defence of a pertinent one. Argumentative Evaluation distinguishes these dimensions and produces a reasoned assessment of each AI model participating in the debate, systematically including a charitable reading before any critical finding.
This mode is suited to debates with high argumentative stakes — scientific controversies, legal debates, strategic negotiations, philosophical dialogues where the rigour of reasoning matters as much as the conclusion. It is valuable for users who must defend a position publicly and want to identify the argumentative weaknesses in their own material before a perceptive counterpart does.
Source Verification
Source Verification audits, as a priority, the factual claims that carry the argument: empirical premises, figures, attributions, studies, mechanisms presented as established, or cases used as evidence. Through actual web research, it checks them against accessible and traceable sources. It does not judge the value of the ideas, but the strength of their factual support.
This verification concerns checkable statements, not theses, interpretations, or value judgements, which no external source can settle and which the mode explicitly excludes from its scope. For each selected claim, it produces a graduated verdict — confirmed, partially correct, not found, not verified, out of scope — together with an explanation and the sources consulted. The result rarely reduces to a binary opposition between true and false: a source may support the general direction of a claim while diverging on its scope, figure, date, or attribution.
This mode is suited to debates with high factual density — scientific controversies, technical dossiers, or subjects where the argument relies heavily on bibliography or quantitative data. It can distinguish an attested figure from one that is merely plausible, flag an overbroad attribution or an untraceable reference, and do so before uncertain claims are relied upon publicly.
How to Choose a Mode
Several questions can guide the choice.
What is the nature of the session to be analysed? An exploratory or creative session that has opened more avenues than it has closed is often best read through Emergence Analysis — which identifies what appeared during the session without having been announced — or Integrative Synthesis, which reconstructs its progression in a coherent form. A session centred on a contested question, where several defensible positions remain opposed without any one prevailing, calls rather for Tension Mapping, which identifies what resists reconciliation, or Meta-Analysis, which reveals what structures the disagreement without the participants’ awareness. A session with high argumentative stakes, where the quality of reasoning matters as much as the content of the conclusions, calls for Argumentative Evaluation.
What kind of knowledge is the analysis meant to produce? If the aim is a directly usable text — for citation, sharing, or presentation to others — Integrative Synthesis is the natural mode: it reconstructs the progression of the session in a readable, reusable text. If the aim is to bring out the presuppositions, biases, and blind spots that structured the session, Meta-Analysis is more appropriate. If the aim is to identify directions worth exploring next, Horizon of Possibilities opens onto what follows rather than closing with a balance sheet. If the aim is to interrogate not the positions themselves but the vocabularies that made them possible, Critical Archaeology traces the discursive conditions that shaped the frame of the debate.
Should several modes be combined? Nothing prevents it, and some combinations produce a richer reading than any one mode used in isolation.
Three combinations are particularly useful. Integrative Synthesis + Tension Mapping suits sessions that need an operational conclusion: the first reconstructs the debate’s progression and points of arrival, while the second keeps persistent disagreements visible. Together, they provide an articulated reading of the positions reached and the tensions that remain open.
Meta-Analysis + Argumentative Evaluation + Source Verification suits sessions with critical stakes: the first identifies the implicit axioms that shaped the exchange, the second evaluates the rigour with which positions were sustained within that frame, and the third audits verifiable claims using web research. Together, they cross-examine the frame, the argumentation, and its factual support.
Emergence Analysis + Horizon of Possibilities suits exploratory sessions: the first identifies what the debate actually brought forth, while the second formulates extensions that the debate has now made possible. Together, they turn the session into one stage of a longer investigation, distinguishing what genuinely emerged from what has become formulable as a result.
Other combinations are of course possible, and experience allows each user to discover those that best fit their own practice. The general principle is simple: the modes do not replace one another; they complement one another. Depending on the objective, two modes may be enough; three or four can bring more perspectives into play when the session warrants it.
A Progression of Use
For someone discovering Metamorfon, the range of eight modes may initially seem daunting. A natural progression makes it possible to enter the system without having to choose among all eight from the first use.
The simplest entry point is Integrative Synthesis. It produces an immediately readable result — a coherent text that reconstructs the progression of the session — and suits most sessions without any special adjustment. For a first analysis, it provides a reliable point of reference.
Once familiar with what Integrative Synthesis produces, users can begin exploring modes that ask different questions of the session. Meta-Analysis is probably the second mode to try because it reveals aspects that the Synthesis does not address — not what was said, but what structured what was said. The contrast between the two modes often conveys, more effectively than any description, what it means to have several perspectives on the same session.
The other modes can then be explored as needs become clearer: Tension Mapping for contested questions, Argumentative Evaluation when preparing a public position, Critical Archaeology when there is reason to suspect that the vocabularies themselves are shaping thought, Emergence Analysis when conducting conceptual research, Horizon of Possibilities when looking for extensions that the debate itself has made formulable, and Source Verification when the factual basis of claims made during the session needs to be checked.
This progression is not prescriptive. Some users immediately find a preferred mode and keep returning to it; others explore more systematically. Metamorfon accommodates both approaches.
For each mode, a dedicated methodological article develops its principles, analytical framework, use cases, and limits in depth. This overview page is a point of entry; the foundational articles develop each approach further.
Try Metamorfon with a 14-day free trial — no commitment required.
Free Trial — 14 days
€0 today, then €30/month incl. VAT
14 days free trial (card required)
Automatically converts to monthly billing after the trial
Cancel anytime