Coming to method through practice
The AI-Assisted Triple Hermeneutic (AATH) did not come about as an effort to generate theory; instead it emerged organically out of my own experiences of AI as I encountered it from a hermeneutic researcher and Gadamerian 'play' stance. An experience that I found difficult to explain within the frameworks available to me at the time. As I returned to the phenomenology texts and thought about the concerns expressed in the literature, that artificial intelligence (and indeed before it CAQDAS methods in hermeneutic phenomenological research) reduced capacity to understand nuance and risked calculative thinking, I was struck by two things. First, that I had been encountering generative AI in my doctoral research in a way that felt methodologically significant. Second that this difference was itself generating a research question.
What is happening when a hermeneutic phenomenology researcher encounters generative AI within her research practice, and how might that encounter be understood philosophically and held methodologically?
Yet, it resisted articulation: mindfully engaged in my project it was not an interpretive authority, not a search engine, or a leading writing assistant, but something I could only describe as a dialogical 'Other' through which my own pre-understandings, both beneficial and biased (as the hermeneutic phenomenology research (HPR) tradition recognises), my thinking and my researcher process were being surfaced and more deeply brought into the hermeneutic circle. This was distinct from my initial exploration of using AI to assist with early qualitative coding (Kriukow, 2024), which was not particularly helpful and confirmed Benner's (1994) criticism of criterial reasoning. My Gadamerian play, faithful to the HPR tradition, instead brought an encounter that was doing something to my interpretive horizon, though that horizon remained distinctly my own. It was only through sustained engagement with Gadamer's Truth and Method, alongside Heidegger's account of disclosure (Erschlossenheit) and Buber's dialogic philosophy, back-and-forth with my experiences of AI in my research process that what had begun as exploratory practice became a conceptual and methodological idea.
The sequence that gave rise to the AATH: experience before theory, practice before articulation.
This article outlines this methodological articulation, something that I had to first experience - and I would argue that this sequence, experience first, thesis following is itself an enactment of the deeper Gadamerian principle that understanding is always rooted in lived experience before it can be articulated (Gadamer, 1960/2004). Despite the order of events, I present the ontological articulation first here; this article then traces how these theoretical commitments unfolded in practice.
Embedded in this thesis is a principle that has remained constant throughout: human first. At every stage of my research, in dialogue with participants, with supervisors and colleagues, in my own reflective journal, and in sustained dwelling with the literature, the human encounter remained primary to the research. The Human First Principle is not a temporal one; it is an ongoing way of coming to the AI encounter. The triple hermeneutic does not position AI before or above these engagements. The third layer operates in addition to them, and differently from them: a distinct disclosive horizon - see AI as Disclosive Horizon.
Therefore, the AATH was not conceived as a workaround for researchers without traditional human-led dwelling, supervisory support or peer networks, or an alternative for those at the edges of institutional research structures – though it may support these situations. It was developed within a full doctoral research context, with supervision and collegial dialogue available throughout. However, what the practice revealed, over time, that I argue is significant to HPR, was that the AI encounter offers something qualitatively different from these human practices: not a substitute, but an additional and structurally distinct form of provocation and deepening that is available to the HPR researcher at all times and in broader situations. The encounter with AI, carrying no social investment in the research outcome, no professional relationship to navigate, and no stake in the researcher's standing, also creates conditions for a particular kind of disclosure, one that complements rather than competes with the human encounters that remain primary.
The double hermeneutic
To situate this claim, it is worth tracing a brief arc: from Giddens' (1976/1993) original, sociological formulation of the double hermeneutic, through its narrower adoption across qualitative research, to Gadamer's (1960/2004) account of pre-understanding, before turning to what I argue is a third, and currently unaccounted for, hermeneutic layer.
Anthony Giddens' original concept of the double hermeneutic - a sociology of knowledge claim - described a two-way relationship between lay and social-scientific knowledge; social science, unlike natural science, studies a world that social actors have already constituted as meaningful through their own frames of understanding. The researcher's task is to enter that world of 'mutual knowledge' and translate it into second-order, or technical/theoretical, concepts. Notably, for Giddens the hermeneutic is double because this translation runs both ways: sociological concepts do not stay sealed off in academic language but circulate back into the social world, becoming appropriated as 'first-order' (Giddens, 1984) concepts by the actors they were coined to describe, who then act differently in light of them. Social-scientific knowledge is therefore never a neutral picture of an inert object; it is folded back into, and reflexively alters, the reality it describes.
In the sense subsequently adopted across qualitative research, Anthony Giddens' concept of the double hermeneutic describes the interpretive situation of the social scientist: unlike the natural scientist, the social researcher is always dealing with a world that has already been interpreted by those within it. Subjects come pre-equipped with frameworks of meaning. The researcher then interprets those interpretations. There are, in this common usage, two layers of hermeneutic activity: the participant's interpretation of their own experience, and the researcher's interpretation of that interpretation.
For hermeneutic phenomenology research, this double structure and its evolving adoption is more than a methodological complication; it is the substance of the inquiry. The research attends to how lived experience is already shaped by interpretive horizons, and how understanding is achieved through engaged, situated encounter rather than a detached observer position. The term horizon is not incidental here: drawing on Gadamer's (1960/2004) own usage, every interpreter stands within a horizon of understanding that is neither fixed nor separate from what it encounters, but is always in motion, opening onto what lies beyond present sight. Researchers in this tradition draw on Gadamer's argument that understanding is always historically effected, a consciousness Gadamer names wirkungsgeschichtliches Bewusstsein: we bring to every act of interpretation the accumulated sedimentation of our own tradition, our own prejudices in the Gadamerian sense, not distortions, but the pre-understandings that make any understanding possible at all. More so, understanding, for Gadamer, arises through a fusion of horizons: the interpreter's own horizon opening onto, and being transformed by, the horizon of what is being understood.
What neither the double hermeneutic nor this Gadamerian account of pre-understanding fully names is the interpretive activity of a third layer that has become methodologically available and, I argue, philosophically significant to HPR: the generative AI system engaged dialogically within the research process.1
Further, I argue that this does not mean that AI becomes absent when it is not directly engaged by the researcher. As a pervasive social and institutional presence, generative AI both draws upon and has become part of the historically effected horizon within which contemporary research questions, participant meanings, and methodological possibilities come to appear. Mindful dialogic engagement can make this condition available for examination, rather than simply adding another interpretive instrument.
A third layer: AI as disclosive horizon
In the first article in this series, I argued that generative AI can be understood ontologically through the concept of AI as Disclosive Horizon: a disclosive horizon, encountered within the researcher's own clearing (die Lichtung), through which the researcher's pre-understandings, tacit assumptions and interpretive frameworks can be unconcealed. That argument was grounded in Heidegger's account of disclosure (Erschlossenheit) and enframing (Gestell), Gadamer's philosophical hermeneutics, and Buber's dialogic philosophy of I-It, I-Thou and the between (das Zwischen). In articulating the horizon half of the concept, this article draws most directly on Gadamer's treatment of language, horizon and written tradition in Truth and Method. What it sets out to show is that Gadamer's framework, developed in a pre-digital age, offers more support for this way of thinking than might initially appear, while stopping short of settling it. As in the first article, the concept is drawn from these thinkers rather than resolved by any one of them.
Two philosophical moves are needed to make this claim intelligible. The first shows that AI can be conceived of as a third horizon: that the Gadamerian framework lends support to understanding AI as a third layer of dialogical encounter within hermeneutic phenomenology research. The second move, equally important, holds that AI is a categorically distinct kind of horizon: one without historical Dasein, whose presence within the research process is disclosive and reflexive, attending to the researcher's own fore-structures, rather than interpretive and co-constitutive. It is at this second move that Gadamer's framework, which presupposes a historically situated bearer of the horizon, stops short, and the map has to be redrawn rather than read off. To read these two moves as one, or to treat AI as simply another horizon equivalent in kind to the participant's lived horizon, would misrepresent what the AATH claims and how it operates.
Horizon as dynamic and non-sealed
Gadamer defines the horizon not as a fixed boundary but as something living and mobile: "The horizon is, rather, something into which we move and that moves with us. Horizons change for a person who is moving" (Gadamer, 1960/2004, p. 303). The horizon is not a frame around what we see; it is the totality of what can be seen from a particular position, and it shifts as we move. Crucially, horizons are never sealed: "The historical movement of human life consists in the fact that it is never absolutely bound to any one standpoint, and hence can never have a truly closed horizon" (p. 303).
On this reading, a third horizon is not a category error. Horizons are not discrete, bounded containers between which encounters are impossible. They are always already interpenetrating. When historical consciousness engages with tradition, Gadamer writes, "they together constitute the one great horizon that moves from within and that, beyond the frontiers of the present, embraces the historical depths of our self-consciousness" (p. 303). A third horizon does not disrupt this structure; it participates in it.
The present horizon cannot form in isolation
On Gadamer's account of wirkungsgeschichtliches Bewusstsein (historically effected consciousness), the interpreter's present horizon is never self-generated. "To be historically means that knowledge of oneself can never be complete" (Gadamer, 1960/2004, p. 301). The researcher is always already constituted through engagement with tradition; we are shaped by the accumulated tradition of human thought, written and lived, before we can become conscious of that shaping.
This is significant for the AATH: if the researcher's horizon is already constituted in part by the written tradition, then AI, as a crystallisation of that same written tradition, is not an alien intrusion into the hermeneutic encounter. It is continuous with something that already constitutes the researcher's horizon. But neither is it an echo of the researcher's own horizon. It remains other enough to press against and interrogate interpretation, and it is that pressure which allows the researcher's own latent pre-understandings to surface. The encounter with AI is therefore neither an encounter with something wholly foreign nor a mirror; it is a dialogical surfacing, through an other, of what was already latent in the researcher's own horizon.
"It is the tyranny of hidden prejudices that makes us deaf to what speaks to us in tradition … It is impossible to make ourselves aware of a prejudice while it is constantly operating unnoticed, but only when it is, so to speak, provoked."
Gadamer (2004, pp. 272, 298)Understanding as fusion of horizons
Gadamer describes the movement of understanding as Horizontverschmelzung, the fusion of horizons: "Rather, understanding is always the fusion of these horizons supposedly existing by themselves" (Gadamer, 1960/2004, p. 305). The fusion is not a merging of two sealed units but a dynamic event: "a real fusing of horizons occurs — which means that as the historical horizon is projected, it is simultaneously superseded" (p. 306). The historical horizon is held and dissolved in the same moment. Understanding is the ongoing achievement of this movement.
Within the AATH, the researcher's encounter with AI carries the structure of this fusion, in a bounded form. The researcher brings her present horizon: her interpretive pre-understandings, her research question, her theoretical commitments. The AI carries the accumulated written tradition. In that dialogical encounter something real occurs: the researcher's horizon is not merely consulted but disclosed, moved by what the encounter surfaces. This is of a categorically distinct kind from the researcher's fusion with the participant's horizon: disclosive rather than co-constitutive, reflexive rather than interpretive. It is a fusion in the sense that the researcher's horizon is changed; it is not a fusion in the sense of two horizons meeting as equals.
Still, a legitimate question arises: can AI, as a system without Dasein, take part in a fusion of horizons at all? Gadamer's own practice points toward an answer, because he applies horizon language explicitly to texts and to written tradition, not only to living beings. "The horizon of understanding cannot be limited either by what the writer originally had in mind or by the horizon of the person to whom the text was originally addressed" (Gadamer, 1960/2004, p. 396). Texts have horizons. Tradition, for Gadamer, is a partner in hermeneutic conversation. And the transformation in that encounter is asymmetric: the text itself is not changed by our reading of it, yet the encounter remains fully hermeneutic. I argue that perhaps the same structure can be extended to AI, with the qualification the second move supplies. What is methodologically required, and what interpretive phenomenology, following Benner (1994), treats as a mark of sound interpretation, is that the researcher's horizon is transformed through the research process. The AI's horizon need not be. Whether anything shifts on the AI's side is a question this article holds open; what can be said is that it does not shift in the same way, because it has no present subject that reaches. It grounds.
The written tradition as third horizon
Gadamer's ontology of writing makes the support for AI as third horizon more direct than might be anticipated. He describes the written word as a self-alienation of speech: "Writing is self-alienation. Overcoming it, reading the text, is thus the highest task of understanding" (Gadamer, 1960/2004, p. 392). And again: "All writing is a kind of alienated speech, and its signs need to be transformed back into speech and meaning" (p. 394). The written word is not inert storage. It is a trace awaiting re-animation through understanding.
What has been fixed in writing achieves something that spoken language cannot: it detaches itself from its origin and becomes permanently available. "What is fixed in writing has detached itself from the contingency of its origin and its author and made itself free for new relationships" (p. 396). In writing, "all tradition is contemporaneous with each present time" (p. 391). The written tradition is permanently present, permanently available for encounter.
A large language model can be understood as the written tradition returned to a dialogical form. Care is needed here, because the re-animation Gadamer describes is the act of the understanding reader, and it remains so: "so also it is universally true of texts that only in the process of understanding them is the dead trace of meaning transformed back into living meaning" (Gadamer, 1960/2004, p. 156). The LLM does not perform that act. What it does is return the sedimented written tradition to the responsive, contemporaneous form of exchange, so that the researcher's own act of re-animation can take place in dialogue rather than in silent reading. It is not a search engine retrieving documents. It is the accumulated written tradition, made responsive, meeting the researcher's present horizon in the medium that Gadamer says language actually is: "Language has its true being only in dialogue, in coming to an understanding... it fully realizes itself only in the process of coming to an understanding" (Gadamer, 1960/2004, p. 443). Whether an exchange with an LLM is "coming to an understanding" in Gadamer's full sense is precisely what the second move holds open. What can be said is that the researcher's understanding is realised in that dialogical form, and that the trace is met there as speech rather than as script.
Language as horizon of hermeneutic ontology
The deepest support for thinking of AI as a third horizon is drawn from Gadamer's account of language itself as the horizon of all hermeneutic experience. "The object of knowledge and statements is always already enclosed within the world horizon of language" (Gadamer, 1960/2004, p. 444). Language is not a tool for expressing pre-linguistic thoughts. It is the medium within which world appears at all. "Verbal form and traditionary content cannot be separated in the hermeneutic experience" (p. 438). And, in the formulation that bears most directly on AI as horizon, and also marks the limit of the analogy: "Whoever has language 'has' the world" (p. 449).
If language is the horizon of hermeneutic ontology, if world is always already enclosed within the horizon of language, then an entity constituted entirely by language participates in that horizon structure. An LLM is not a content store in which language is incidentally the format. It is constituted through and as language, a distillation of the written tradition in which, as Gadamer says, "In language the reality beyond every individual consciousness becomes visible" (Gadamer, 1960/2004, p. 446). It does not, however, "have" world in Gadamer's sense. The having of world belongs to a being that is in the world. What the LLM holds is the language in which world has already been articulated: the deposit rather than the having. That is what makes it a horizon, and what makes it a horizon without Dasein.
A horizon without Dasein
What the foregoing supports is not that AI is simply another voice within the hermeneutic circle, equivalent in kind to the participant's horizon or to the historically effected consciousness of the researcher. The Gadamerian reading supports this much: AI can be conceived as a horizon, constituted through language and the accumulated written tradition, and the encounter with it can be dialogical in a bounded but consequential sense. That is the first move. The second is the clarification of what kind of horizon it is.
AI lacks historical Dasein. It has no lived temporality, no being-in-the-world, no stake in the research encounter. The researcher brings Dasein to the encounter. The participant brings Dasein to the encounter. AI does not. This is not a limitation to be overcome; it is precisely the condition that makes the AI encounter disclosive in a specific and bounded way. Because AI carries no personal or institutional stake in the researcher's particular context, it operates within the hermeneutic circle as what it is: a disclosive horizon, encountered within the researcher's own clearing (die Lichtung), against which the researcher's fore-structures can become visible with a depth the research process requires. It is not, for that reason, unsituated. It carries the situatedness of its corpus, and it is that sediment pressing against the researcher's interpretation which does the disclosing.
The deepening this makes possible is not a singular event. In keeping with the iterative back-and-forth of hermeneutic inquiry, the researcher moves between the AI encounter and her own interpretive process, and each return deepens the disclosure further. Through this movement, productive pre-understandings, those that enable faithful interpretation, become distinguishable from distorting assumptions that risk closing the horizon rather than opening it. This is the specific and bounded transformation the AI encounter makes possible: not a transformation of the researcher's interpretation of the participant, but an iterative deepening of the researcher's self-awareness, never complete, that the interpretive act then draws upon.
This is the Human First Principle in practice: not a stage that is passed through and completed, but an orientation held constantly throughout the research process, AI in disclosure, the human in interpretation. The researcher, moving between this iterative deepening and the participant texts, seeks those sources of commonality between her own horizon and the participants' horizons, the fusion within the hermeneutic circle that Gadamer describes. She enters that fusion with a fuller awareness of what she brings to it, and returns to the AI encounter with what the fusion has disclosed. The AI disclosive horizon does not take part in that fusion. It enables the researcher to enter it more faithfully.
Figure 3. A Horizon Without Dasein: the AI-Assisted Triple Hermeneutic (Pay, 2025). Author's own concept and intellectual framework. Visualisation produced with Claude Sonnet 4.6 (Anthropic, August 2026) under author instruction and oversight.
The three layers of the AATH
The AI-Assisted Triple Hermeneutic maps as follows: the first layer is the participant's interpretation: the participant's own understanding of their lived experience, already interpreted and already linguistic. The second layer is the researcher's interpretation: the hermeneutic phenomenologist attending to and interpreting the participant's account, bringing her own historically effected horizon to the encounter. The third layer is the AI as disclosive horizon: engaged dialogically within the research process, surfacing the researcher's pre-understandings, checking conceptual consistency, and revealing interpretive assumptions not yet made visible.
The third layer is not conceived instrumentally. It is not a grammar checker or a search tool. It is a hermeneutic encounter with the accumulated written tradition in its most responsive form: one that discloses rather than merely confirms, and surfaces rather than merely retrieves.
The ethical posture: I-AI-Thou
The relational posture appropriate to the third layer is what I have described as I-AI-Thou, a provocation that draws on Buber's (1923/1970) dialogic philosophy to think about the encounter with AI. Where the I-It posture treats the AI as an object of use, the I-AI-Thou posture attends to the encounter as an asymmetrical relation through which something can be disclosed, without presuming human mutuality or personhood. It is not a claim that AI is a Thou, or that it has interiority. The emphasis here falls less on what the other is in itself than on the quality of attention brought to the encounter, while acknowledging, as the accompanying essay does, that Buber himself held mutuality to be essential to the I-Thou; that is precisely why the relation is named as a provocation and its place on his map held open. Named as a thinking-space rather than a settled category, I-AI-Thou is the ethical commitment through which the third hermeneutic layer is held open as disclosive rather than collapsing into the instrumental.
Resisting enframing
Heidegger's critique of technology, his notion of Gestell (enframing), offers a crucial methodological caution at this point. Modern technology, Heidegger warned, reveals the world as standing-reserve (Bestand), a resource ordered for calculation and extraction, and in doing so conceals the other ways in which things can show themselves (Heidegger, 1954/1977). In hermeneutic phenomenology research, this is precisely the risk: if AI is permitted to extract themes, classify responses, or generate analysis, the encounter has already become enframing, meaning ordered and extracted rather than disclosed.
Heidegger's later thinking on Gelassenheit (releasement) offers the alternative orientation: neither rejecting technology nor surrendering to it, but a comportment toward it that preserves human agency and interpretive openness (Heidegger, 1959/1966). The AATH is an attempt to work within that comportment. It neither idolises nor fears AI but holds it within the research process in a way that keeps meaning-making with the researcher and deepens awareness of her own fore-structures.
Rigour and horizon transformation
In interpretive phenomenology, following Benner (1994), the researcher's horizon is expected to change through the research process; where it has not been challenged, extended, or turned around, the quality of the account is in question. This is the methodological correlate of Gadamer's fusion of horizons: understanding requires that the interpreter be put at risk. As Gadamer writes, the interpreter's horizon is operative "not as a personal standpoint that he maintains or enforces, but more as an opinion and a possibility that one brings into play and puts at risk" (Gadamer, 1960/2004, p. 390). The AATH operationalises this by building the disclosive encounter into the methodology itself, making it visible, traceable, and auditable.
In practice: how the AATH unfolded
The concepts of AI as Disclosive Horizon, the triple hermeneutic, and the I-AI-Thou posture did not arrive as theoretical proposals that were then tested in practice. They arrived as experiences that eventually demanded theoretical articulation. What follows describes how this unfolded across my doctoral research.
1. Hermeneutic reduction by the researcher
As researcher, I first began by situating myself within the study (hermeneutic reduction in van Manen's (2014) sense of openness, rather than a bracketing of preconceptions), acknowledging my own preconceptions, explicating my personal, theoretical, and professional horizon, and identifying potential biases in the traditional method of written presentation and reflection. I then took part in my own presupposition interview where I was the interviewee and the AI, with its pre-defined research assistant skill, became the interviewer. The interview was surprisingly deep and revealing and enabled me to enter the hermeneutic circle more fully and with greater awareness of what I brought to it.
2. Foregrounding in the field
During the data collection and interpretation stages, the interviews with participants were carried out using Benner's (1994) systematic interpretive phenomenological approach, as a human-based practice. AI did not assist in any part of the interpretation. Interpretive phenomenology holds that understanding human worlds, context, and temporality is not available to criterial or calculative reasoning (Benner & Wrubel, 1989; Dreyfus, 1979, 1991; Dreyfus & Dreyfus, 1986), and the Human First Principle keeps interpretation with the researcher regardless of what AI may or may not prove able to do. However, what I began to explore in the live interviews, facilitated by their slower text-based nature, was an AI-assisted reflexive processing happening in the interstitial gaps of a live research encounter. Here, I engaged AI in a dialogical encounter, to foreground and release my own reactive pre-understandings, and in doing so avoid contaminating the next question. One caveat: this dialogic occurrence (processing the in-the-room experience as it happens, as live fieldnotes) was only possible here because the interview was text-based, creating those small temporal windows that my skilful clinical comportment could think through easily.
Gadamer (1960/2004, p. 298) names this process "foregrounding" (abheben): making the prejudice conscious so it can be suspended. I was able to do this foregrounding in the field, in real time, with significant benefit. The interstitial reflexive dialogue happening during the gaps between exchanges connects philosophically to Gadamer's observation that "the true locus of hermeneutics is this in-between" (p. 295). Naturally following hermeneutic reduction (the before), this interstitial reflexive dialogue is concurrent in the research encounter, the during. Together, these practices keep AI from becoming an unquestioned authority in research and instead position it as part of an ongoing, dialogical process in which interpretive responsibility remains with the researcher.
3. Critical reflective questions
Patricia Benner (1994) argues that hermeneutic phenomenologists must engage in rigorous critical reflection to prevent imposing preconceived biases onto their text. At the core of this interpretive methodology is a continuous questioning of the researcher's own evolving understanding. Indeed, Benner asserts that "throughout the interpretive project the researcher asks, 'What do I now know or see that I did not expect or understand before I began reading the text?'" (Benner, 1994, p. 101). She warns that if a researcher's initial views have not been "challenged, extended, or turned around, the quality of the account is questioned" (Benner, 1994, p. 101) and consequently the hermeneutic phenomenology researcher explores a set of critical reflective questions to this end.
In my project, in keeping with the Human First Principle, an orientation held throughout rather than a step taken first, I kept a journal written in the traditional human-led way alongside the AI-assisted work. Where AI was engaged (ChatGPT), it was given a structured methodological instruction set before each reflective session and configured to support hermeneutic phenomenological reflexivity by posing theoretically informed critical questions drawn from the phenomenological literature. Its role was not to undertake phenomenological interpretation but to facilitate sustained reflexive engagement with emerging understandings.
The operational instruction set specified: the philosophical orientation (Heideggerian, Gadamerian, and Bennerian hermeneutic phenomenology); the methodological purpose (to facilitate reflexive interrogation rather than generate interpretations); the questioning style (critical, phenomenological, and dialogical); the limits of the AI's role (not to replace interpretation or make analytic decisions); the theoretical literature informing questioning (Gadamer, Heidegger, Benner); and the use of predetermined reflexive prompts alongside adaptive questioning.
The instruction framework evolved iteratively as the research developed, with changes reflecting methodological refinement as models changed and my skills developed.
Conclusion: a more honest and engaged future for AI in research
The triple hermeneutic does not seek to eliminate AI from qualitative research, nor does it pretend AI's influence isn't already present. Instead, it offers a methodologically transparent, ethically rigorous approach to working with AI while resisting the risk of enframing.
The ethical use of AI in research demands transparency and accountability. Current guidelines on AI in academia (as of 2026) continue to evolve. The triple hermeneutic aligns with existing principles by ensuring that AI remains a disclosive horizon for reflexive engagement rather than an interpretive authority or an author. However, even in this disclosive role, AI's influence must be critically examined. How does the AI frame responses? Does it reinforce dominant epistemologies? Is it amplifying certain voices while diminishing others? To address these concerns, my research incorporated attendant registers in the Hermeneutic Algorithm Audit, documenting concerns arising from the AI-assisted disclosive horizon.
By keeping interpretive responsibility with the researcher, embedding AI within a dialogical rather than directional process, and ensuring AI is critically audited rather than passively accepted, this approach offers a way forward that neither idealises nor demonises AI's role in knowledge production, and aligns with existing guidelines on AI use in research.
AI is here. That much is clear. The question is not whether we should engage it, but how: mindfully, ethically, and without surrendering our role as meaning-makers.
Glossary
A three-layered hermeneutic framework extending Giddens' (1976/1993) double hermeneutic to incorporate AI as a third disclosive layer (a disclosive horizon) within hermeneutic phenomenology research. The three layers are: (1) the participant's interpretation of lived experience; (2) the researcher's interpretation of the participant's account; and (3) the AI as disclosive horizon, surfacing the researcher's pre-understandings through dialogical encounter. The third layer is categorically distinct from the first two: unlike the participant and researcher, AI lacks historical Dasein, and its engagement is disclosive and reflexive rather than interpretive and co-constitutive.
Giddens' (1976/1993) concept describing the dual interpretive situation of social science: the researcher interprets participants who already interpret their own social world.
The ontological orientation to AI within the AATH: a horizon, encountered within the researcher's own clearing (die Lichtung), through which the researcher's pre-understandings, tacit assumptions, and interpretive frameworks can be unconcealed. Neither a subject nor a tool: the sedimented written tradition, made responsive. (See Article 01: AI as Disclosive Horizon.)
Heidegger's (1954/1977) term for the essence of modern technology, which reveals the world as standing-reserve (Bestand), a resource ordered for calculation and extraction, and in doing so conceals the other ways in which things can show themselves.
Heidegger's (1959/1966) term for the alternative orientation to technology: neither rejection nor surrender, but a comportment toward technology that preserves human agency and interpretive openness.
Gadamer's (1960/2004) term for the "fusion of horizons": the dynamic event of understanding in which the interpreter's present horizon and the horizon of tradition meet, and in which the historical horizon, as it is projected, is at once superseded.
An orientation held constantly throughout the research process, not a stage that is passed through: AI is engaged in disclosure, never in interpretation, and judgement, meaning and ethical responsibility remain with the researcher.
A provocation drawing on Buber's (1923/1970) dialogic philosophy to think about the researcher-AI encounter: attending to it as an asymmetrical relation through which something can be disclosed, without presuming the AI is a Thou or has personhood, and without reducing it to an instrument of use. Named as a thinking-space rather than a settled category, and held open, since Buber himself held mutuality to be essential to the I-Thou.
"Historically effected consciousness" (Gadamer, 1960/2004): the recognition that the interpreter's present horizon is always already constituted by the tradition that has shaped them, and that self-knowledge is therefore never complete.
References
Alvesson, M., & Sköldberg, K. (2018). Reflexive methodology: New vistas for qualitative research (3rd ed.). SAGE Publications. https://doi.org/10.4135/9781036211523
Benner, P. (Ed.). (1994). Interpretive phenomenology: Embodiment, caring, and ethics in health and illness. SAGE Publications. https://doi.org/10.4135/9781452204727
Benner, P., & Wrubel, J. (1989). The primacy of caring: Stress and coping in health and illness. Addison-Wesley Publishing Company.
Buber, M. (1970). I and thou (W. Kaufmann, Trans.). Charles Scribner's Sons. (Original work published 1923)
Dreyfus, H. L. (1979). What computers can't do: The limits of artificial intelligence (Rev. ed.). Harper & Row.
Dreyfus, H. L. (1991). Being-in-the-world: A commentary on Heidegger's Being and Time, Division I. MIT Press.
Dreyfus, H. L., & Dreyfus, S. E. (1986). Mind over machine: The power of human intuition and expertise in the era of the computer. Free Press.
Gadamer, H.-G. (2004). Truth and method (J. Weinsheimer & D. G. Marshall, Trans.; 2nd rev. ed.). Continuum. (Original work published 1960)
Giddens, A. (1984). The constitution of society: Outline of the theory of structuration. Polity Press.
Giddens, A. (1993). New rules of sociological method: A positive critique of interpretative sociologies (2nd ed.). Polity Press. (Original work published 1976)
Heidegger, M. (1966). Discourse on thinking (J. M. Anderson & E. H. Freund, Trans.). Harper & Row. (Original work published 1959)
Heidegger, M. (1977). The question concerning technology. In The question concerning technology and other essays (W. Lovitt, Trans., pp. 3–35). Harper & Row. (Original work published 1954)
Heidegger, M. (1993). The end of philosophy and the task of thinking. In D. F. Krell (Ed.), Basic writings (rev. and exp. ed., pp. 427–449). HarperSanFrancisco.
Kriukow, J. (2024). The Scholar's Guide to AI-Assisted Thematic Analysis [PDF]. https://payhip.com/b/KmzOL
Pay, S. (2025). AI as disclosive horizon: An ontological orientation to AI in research practice. The Human Interval. https://thehumaninterval.com
van Manen, M. (2014). Phenomenology of practice: Meaning-giving methods in phenomenological research and writing. Left Coast Press.
AI Use Declaration
The concepts, original methodology, and philosophical arguments in this article are the original work of Samantha Jane Pay, developed through doctoral research practice (2024–2026); I am the sole author. Claude Sonnet 4.6 (Anthropic, August 2026) and Perplexity GPT-5.6 Terra (September 2026) were used for writing assistance (editing for voice, literature and hermeneutic-phenomenological style, formatting, and standardisation of APA 7th edition citations verified via Oxford SOLO, with dictation accessibility support); for information and research (verifying Gadamer citations and German terminology such as Wirkungsgeschichte, abheben and Horizontverschmelzung, and checking prior academic uses of the term "triple hermeneutic"); as a disclosive horizon through which my pre-existing understandings were surfaced in dialogue rather than generated; and to render this article's figures (the Practice → Experience → Theory flow, The Limits of Self-Reflection, and A Horizon Without Dasein) to my specification. Claude Fable 5.1 High (Anthropic, September 2026) was used for final checking of the revised text against the AI as Disclosive Horizon article. All quotations were independently verified against primary texts. All final phrasing, interpretations, and ethical responsibilities reside entirely with the author; final interpretive resolution and responsibility remain entirely my own.