Scoping Review Maps Husserl’s Explicit Role in Artificial Intelligence Research

A new scoping review has systematically charted where one of the twentieth century’s most demanding philosophers shapes modern artificial intelligence research. The review, published in Discover Artificial Intelligence by Hendrik Poschmann and Stefania Centrone, retrieved and screened English-language publications through a reproducible single-source search for work that explicitly anchors AI or robotics research in the phenomenology of Edmund Husserl. The result is a strikingly narrow field: out of 5,594 records retrieved through the Semantic Scholar Academic Graph API on 5 December 2025, only 32 publications, spanning 1990 to 2025, satisfied the review’s stringent criterion of a materially operative Husserlian anchor. That inclusion ratio of just 0.57 percent is itself analytically revealing, the authors argue, because it exposes a vast adjacent literature on machine consciousness, embodiment, and phenomenology in a loose sense, of which only a small subset genuinely engages Husserl’s distinctions rather than borrowing his vocabulary secondhand.

The motivation behind the review is a mismatch that has become increasingly visible as AI systems are credited with perception, understanding, explanation, and even rudimentary agency. A model may be said to understand because it performs well on a benchmark; a robot may be said to perceive because it classifies sensory input; a conversational system may be said to reason because its outputs exhibit discursive coherence. Yet each of these attributions, the authors contend, presupposes an often-unstated view of what it means to relate to a world, sustain meaning over time, and operate within a horizon of relevance. Husserlian phenomenology supplies exactly the missing conceptual apparatus. In his framework, intentionality denotes consciousness as consciousness of something, analysed through the correlation between noetic acts and noematic sense, where an object is intended under a determinate mode of givenness rather than represented as a merely internal item. Constitution names the lawful syntheses through which sense, identity, and objectivity are articulated across changing appearances, while horizonality concerns the co-intended background of further possible givenness.

Crucially, the review treats behavioural and functional success as insufficient evidence of consciousness. The authors identify a classic logical fallacy at work in much AI discourse: the inference that if conscious intelligence can produce behaviour B, and a machine produces B, then the machine is conscious, merely affirms the consequent and is invalid. Human-like performance supports at most a claim of functional similarity. But the invalidity of that inference does not establish the converse modal claim that artificial consciousness is impossible; such a conclusion would require an independent ontological argument. The review therefore remains deliberately neutral on whether artificial consciousness or intentionality is possible in principle, insisting that a categorical exclusion demands its own justification just as much as a premature attribution does. This balanced stance distinguishes the work from both uncritical anthropomorphism and dogmatic denial.

The methodological design followed the PRISMA Extension for Scoping Reviews, using a four-stage pipeline of deduplication, title screening, abstract screening, and full-text assessment. After removal of 67 duplicates, 5,527 records remained for title screening, of which 4,594 were excluded; abstract screening removed a further 875, leaving 58 full texts, of which 26 were excluded at final assessment. Screening was conducted by a single reviewer, a choice the authors acknowledge as a limitation, mitigated by a strict codebook, standardized exclusion reasons, and traceable stage-by-stage decisions documented in a public repository. The eligibility criteria required an explicit Husserlian anchor, either direct reference to Husserl or attributable use of distinctively Husserlian constructs, alongside substantive relevance to AI or robotics. Generic appeals to phenomenology, or reliance exclusively on later traditions such as Heidegger, Merleau-Ponty, or Dreyfus without explicit Husserl anchoring, were excluded. This boundary was a substantive decision, not an arbitrary narrowing: it prevents terms such as intentionality, horizon, embodiment, or lifeworld from being counted as phenomenological merely because they appear in anti-reductionist or situated approaches.

The temporal distribution of the corpus shows a marked recent intensification, with 18 of the 32 included sources appearing from 2021 onward. The sample thus connects earlier debates on symbolic cognition and embodiment with newer questions about explainability, generative AI, and machine consciousness. Substantively, the publications cluster around four recurring domains: embodied AI and robotics, AI epistemology, cognitive architectures, and machine consciousness, with a newer generative-AI cluster unmistakably present but remaining primarily conceptual in character. In this newest cluster, Husserl is invoked above all to diagnose the limits of fluency, authorship, semantic plausibility, and social world-disclosure rather than to guide technical implementation. The most frequently mobilized Husserlian concepts across all domains are constitution, intentionality, horizonality, and the lifeworld, with time-consciousness and intersubjectivity recurring prominently in robotics and machine-consciousness discussions.

When organized by mode of influence, the corpus reveals that conceptual translation is the most frequent contribution. Here Husserl is not invoked merely as an authority but as a source of distinctions rearticulated in AI-relevant terms: reinterpretations of intentionality as structured, temporally mediated aboutness rather than primitive directedness; accounts of perception as constituted rather than passively received; and analyses of embodiment through the distinction between the lived body and the merely physical body. The second major mode is critique, in which Husserl disciplines AI discourse by clarifying what current systems cannot yet be said to do, targeting inflated claims about robot consciousness, shallow conceptions of embodiment, and reductions of meaning to statistical patterning. Agenda-setting publications reorient what AI research should attend to, treating AI as a transformation of the lifeworld, examining virtual personalities as quasi-others without subjectivity, and reframing responsible AI and explainability as problems of practical intelligibility and shared horizons.

The smaller but especially informative category is methodological import and design guidance, which occurs in seven of the 32 corpus publications. The clearest cases concentrate in embodied AI and robotics. Kitamura and Ono proposed one of the earliest architecture-facing contributions, a six-layer consciousness-based robot control system in which anticipation, behaviour inhibition, and emotion-sensitive selection function as organizational principles, with horizonality serving as a justificatory concept for anticipatory architecture. Later work by Tani interpreted retention and protention, Husserl’s terms for the carry-over of just-past experience and the anticipation of what is coming, through context-dependent recurrent dynamics and winner-take-all segmentation, showing that inner time-consciousness can inform a concrete computational interpretation rather than remaining a purely philosophical motif. A paper by Incao, Sciutti, and Gallagher argues that intrinsic temporality should be built into cognitive architectures as a structural condition for minimal self-monitoring and predictive coordination, rather than outsourced to memory modules appended after the fact.

Other contributions bring Husserl into empirical proximity. Chame and colleagues used active inference and an openly available neurorobotics framework to model primary intersubjectivity in second-person interaction, moving further than most texts toward actual operationalization of a Husserl-adjacent idea. Parviainen and colleagues introduced the double body effect, which distinguishes material presence from lived expressivity and explains how robots can elicit genuine affective responses without possessing bodily awareness. Guillermin and Georgeon criticized what they call the isolated-perception paradigm, proposing Artificial Interactionism, which models inputs as outcomes of action-oriented control loops rather than as givens for cognition. Following peer review, a post hoc sensitivity check against PhilPapers recovered nine apparently eligible publications absent from the original retrieval; these were charted separately using the existing codebook and not added to the formal corpus. Methodological import or design guidance applied to three of these nine external publications, a higher descriptive share than in the corpus, which the authors note precludes claiming that the quantitative balance of the retrieved sample is stable.

The review also confronts the changed technical landscape. Dreyfus’s celebrated critique targeted rule-based, explicitly representational symbolic AI, but deep learning, Transformers, and foundation models learn multi-level representations from data and organize sequence processing through attention, dispensing with explicit hand-coding of symbolic rules. Readings of classical AI criticism focused exclusively on explicit rule manipulation cannot by themselves specify what is missing in transformer-based systems, the authors argue, which requires a more differentiated account of how concerns about situated coping and lived embodiment apply. They also warn against equating phenomenological terms with the nearest functional construct: the Husserlian lifeworld, the pregiven world of practical familiarity within which theoretical abstractions become meaningful, is not global workspace theory, and increasing a language model’s context window changes computational coordination without establishing participation in a shared horizon of practice.

The limitations are candidly acknowledged. The review is a reproducible single-source sample, and the nine omitted publications are material relative to a corpus of 32, bearing directly on quantitative claims about constructs, domains, and influence modes. English-language and date restrictions further bound the map, much of the material is theoretical, and neither influence-mode counts nor philosophical engagement measure engineering effectiveness. Sections developing the authors’ constructive proposals, including layered evaluation frameworks distinguishing sensorimotor registration from context-sensitive stabilization of object identity, and research agendas for humanoid platforms testing temporal continuity and self/other discrimination, are explicitly separated from the mapped findings. The review’s enduring contribution lies in its structured account of contributions that a generic phenomenological label obscures: temporality connects philosophical analysis to continuity and anticipation in cognitive architectures, embodiment and intersubjectivity clarify interaction requirements, and horizonality and lifeworld analysis sharpen questions about meaning, evidence, and practical intelligibility, offering AI research a focused way to bring Husserlian analysis into dialogue with its own technical ambitions.

Subject of Research: A scoping review mapping the role of Husserlian phenomenology in artificial intelligence research

Article Title: A scoping review mapping the role of Husserlian phenomenology in artificial intelligence research

Article References: Poschmann, H., & Centrone, S. (2026). A scoping review mapping the role of Husserlian phenomenology in artificial intelligence research. Discover Artificial Intelligence, 6(1), Article 1623. https://doi.org/10.1007/s44163-026-02602-4

Image Credits: AI Generated

DOI: 10.1007/s44163-026-02602-4

Keywords: Husserl, phenomenology, artificial intelligence, scoping review, machine consciousness, embodied AI, robotics, intentionality, cognitive architectures, generative AI, lifeworld, time-consciousness