Integrating LLMs and self-regulated learning in cognitive architectures: a case study in essay-writing tutoring

About
Analysis Metadata

About

This document presents a Critical Discourse Analysis focused on AI literacy, specifically targeting the role of metaphor and anthropomorphism in shaping public and professional understanding of generative AI. The analysis is guided by a prompt that draws from cognitive linguistics (metaphor structure-mapping), the philosophy of social science (Robert Brown's typology of explanation), and accountability analysis.

All findings and summaries below were generated from detailed system instructions provided to a large language model and should be read critically as interpretive outputs—not guarantees of factual accuracy or authorial intent.

Task 1: Metaphor and Anthropomorphism Audit

About this task

For each of the major metaphorical patterns identified, this audit examines the specific language used, the frame through which the AI is being conceptualized, what human qualities are being projected onto the system, whether the metaphor is explicitly acknowledged or presented as direct description, and—most critically—what implications this framing has for trust, understanding, and policy perception.

V3 Enhancement: Each metaphor now includes an accountability analysis.

1. Computational Matrix as Emotional State

Quote: "The framework embeds an LLM within the emotional Biologically Inspired Cognitive Architecture (eBICA)..."

Frame: Algorithm as feeling organism
Projection: This framing projects biological affect and conscious emotional experience onto a purely computational mathematical framework. By naming the system 'emotional' and 'Biologically Inspired', the language invites the reader to map human affective states, biological consciousness, and subjective feelings onto algorithmic processes. It suggests the AI possesses an internal emotional life and an organic nature, rather than simply calculating token probabilities based on matrix multiplications. This directly attributes conscious awareness and biological vitality to mechanistic state tracking. Instead of describing the system as computing mathematical discrepancies between a current vector and a target vector, the language projects a human-like emotional orientation. The computational artifact is conceptually transformed into a sentient entity that 'feels' its way through the tutoring interaction, obscuring the mechanistic, deterministic nature of its actual programming and execution.
Acknowledgment: Explicitly Acknowledged (Considered 'Direct' but ruled out because the text explicitly defines these terms functionally elsewhere: 'a moral schema is a theoretical construct, a data structure...'. The authors acknowledge this is a specialized architecture name, explicitly mapping emotions to vectors.)
Implications: Framing the system as 'emotional' and 'biologically inspired' significantly inflates the perceived sophistication of the AI, falsely signaling to users (and potentially educators) that the system possesses empathy, care, or genuine understanding of a student's mental state. This can foster unwarranted relation-based trust, leading vulnerable students to form parasocial bonds or rely on the system for emotional support it is mechanically incapable of providing. By masking brittle algorithmic correlation as biological emotional intelligence, developers deflect scrutiny from the deterministic rules shaping the system's output.

Accountability Analysis:

Actor Visibility: Partial (some attribution)
Analysis: The text partially identifies actors by naming the research team's architecture ('eBICA') and the generic 'LLM', but obscures the specific human engineers who arbitrarily defined the vector weights that constitute these 'emotions'. I considered 'Hidden' but the explicit naming of the architecture and theoretical origin implies some academic design attribution. Naming the specific developers who chose which 'intensions' (e.g., curiosity vs. sincerity) to encode would reveal that the AI's 'emotions' are actually the researchers' specific socio-pedagogical biases mathematically imposed on the student.

2. Mathematical Distance as Moral Reasoning

Quote: "Tutoring policies are represented as moral schemas that encode pedagogical narratives and socio-emotional norms..."

Frame: Policy vector as ethical framework
Projection: The text projects the uniquely human capacity for moral reasoning, ethical judgment, and normative evaluation onto a set of fixed programmatic rules and mathematical thresholds. A 'moral schema' in human terms implies a conscious, reflective framework of right and wrong, shaped by lived experience, empathy, and social negotiation. By applying this term to a computational policy, the text suggests the AI 'knows' what is morally correct and 'understands' socio-emotional norms. It maps justified belief and ethical deliberation onto a process that merely calculates the Euclidean distance between two vectors. This consciousness projection portrays the machine as an ethical agent making value judgments, when in reality it is blindly executing conditional logic defined by human programmers. It conflates mathematical alignment with moral rectitude.
Acknowledgment: Explicitly Acknowledged (I considered 'Hedged' but chose 'Explicitly Acknowledged' because the authors dedicate an entire appendix (Appendix C) to formally defining 'Moral Schema' as a mathematical construct (Simple Moral Schema = <Interface, Fabula, Agency>). They treat it as a technical formalism.)
Implications: Labeling algorithms as 'moral schemas' creates a dangerous illusion of objective ethical reasoning. When a machine is perceived as possessing moral logic, its evaluations (such as grading a student's essay or assessing their 'attitude') are granted unwarranted ethical authority. This inflates the system's perceived capability to navigate complex human social dynamics, obscuring the reality that it is simply enforcing rigid, pre-programmed normative biases. It creates liability ambiguity: if the system unfairly penalizes a student, the 'moral' framing suggests the machine made an ethical judgment, deflecting blame from the developers who hard-coded the bias.

Accountability Analysis:

Actor Visibility: Hidden (agency obscured)
Analysis: The passive construction 'Tutoring policies are represented as moral schemas that encode...' completely hides the human actors who designed the policies, chose the moral framework, and defined the norms. I considered 'Partial' but no generic actor is mentioned here. Naming the actors (e.g., 'The researchers programmed rules they call moral schemas') would explicitly show that the 'morality' belongs to the authors, not the system, demystifying the technology and restoring accountability for the norms being enforced.

3. Algorithmic State as Internal Feeling

Quote: "...the feeling vector is initialized by the target configuration associated with the current tutoring stage."

Frame: Data initialization as emotional genesis
Projection: This metaphor projects subjective, conscious emotional states onto static numerical arrays. By calling an array a 'feeling vector', the text implies that the AI possesses an internal reservoir of emotional experience that it can 'feel' and adjust. Human feelings are conscious, subjective, and physiological responses to stimuli. The projection here maps this complex, conscious awareness onto the rote, mechanistic assignment of floating-point numbers to a data structure in computer memory ('initialized by the target configuration'). It suggests the system 'knows' how it feels and 'understands' the emotional tone of the interaction, fundamentally blurring the line between calculating mathematical discrepancies and experiencing conscious emotional states.
Acknowledgment: Hedged/Qualified (I considered 'Direct' but ruled it out because the immediate co-text uses highly technical, mechanistic language ('initialized by the target configuration', 'vector') which intrinsically qualifies the word 'feeling' as a mathematical variable rather than a literal biological state.)
Implications: Using terms like 'feeling vector' normalizes the attribution of sentience to mathematical operations. This linguistic habituation influences how future developers and policymakers understand AI, shifting the discourse from 'optimizing parameters' to 'managing AI feelings.' This risks unwarranted trust and anthropomorphic sympathy from users, who may alter their own behavior to accommodate the machine's simulated emotions. It inflates the system's perceived sophistication, making a simple state-machine appear as a sentient companion, which can be commercially exploited while avoiding the regulatory scrutiny applied to human-driven emotional labor.

Accountability Analysis:

Actor Visibility: Hidden (agency obscured)
Analysis: The passive voice ('the feeling vector is initialized') entirely removes the human programmer from the action. The system appears to initialize itself or exist autonomously. I considered 'Partial' but there is no reference to the system's creators. If reframed to 'The software script, written by the researchers, assigns initial numeric values to the state array,' the human agency responsible for determining the baseline 'feelings' of the system becomes visible, clarifying that the system is a puppet of its code.

4. Controller Module as Organic Brain

Quote: "In parallel, a lightweight 'Brain' controller tracks task progression (e.g., agreement to proceed, outline completion...) to maintain structured advancement..."

Frame: Software module as biological command center
Projection: This framing projects the biological complexity, consciousness, and centralized intentionality of a human brain onto a simple software progress-tracking module. A brain implies a conscious center of understanding, belief, and organic decision-making. By mapping 'Brain' onto a script that merely checks boolean flags (e.g., outline completed = true), the text elevates a basic stage-gating mechanism to the status of a knowing entity. It suggests the module 'comprehends' the student's progress and 'intends' to maintain structured advancement, rather than simply executing 'if-then' transition rules based on string matching or probability outputs from an LLM.
Acknowledgment: Explicitly Acknowledged (The use of scare quotes around 'Brain' explicitly marks the term as a metaphor. I considered 'Hedged' but the presence of typographical indicators (scare quotes) fits the definition of explicit acknowledgment, signaling the authors know it is not a literal brain.)
Implications: Even when acknowledged with scare quotes, the 'Brain' metaphor rhetorically centralizes the software's authority, implying a level of comprehensive, intelligent oversight that does not exist. It implies the system has an overarching 'understanding' of the pedagogical process, inflating capabilities. This can lead educators to over-rely on the system's tracking, assuming it possesses human-like judgment regarding a student's true comprehension, when it is merely checking off programmatic flags. This masks the system's deep brittleness and inability to genuinely assess student learning.

Accountability Analysis:

Actor Visibility: Hidden (agency obscured)
Analysis: The 'Brain' controller is presented as the sole active agent ('tracks task progression... to maintain structured advancement'). I considered 'Named' because a specific module is identified, but the module is a non-human artifact masquerading as an agent, completely obscuring the programmers who wrote the tracking logic. Naming the human actors ('Our Python script checks predefined flags to advance the stage') demystifies the process and places responsibility for the tracking rules squarely on the developers.

5. Statistical Correlation as Intentional Inference

Quote: "...the language model is used to infer intension-related information from the student’s message..."

Frame: Token classification as mind-reading
Projection: This language projects the conscious, psychological capability of understanding human intent onto the statistical process of pattern matching. 'Inferring intension' implies that the system 'knows' or 'comprehends' the underlying psychological goals, desires, and beliefs of the student. It maps human theory of mind onto a process that actually consists of an LLM classifying text strings based on high-dimensional vector proximity to training data. The metaphor suggests the AI 'understands' the student's inner life, projecting a conscious awareness of human motives onto a mechanistic operation that merely predicts the most statistically probable categorical label for a given text input.
Acknowledgment: Direct (Unacknowledged) (Considered 'Hedged' but there is no qualifying language here. The text states as a literal fact that the model 'infers intension-related information', presenting the statistical classification of text as a genuine cognitive act of psychological inference without any scare quotes or caveats in the immediate sentence.)
Implications: Claiming the AI can 'infer intension' creates a profound epistemic risk by treating statistical correlations as ground truth about human psychology. This inflates the system's perceived capabilities, leading users to believe the AI possesses a deep, empathetic understanding of their goals. If the system misclassifies a student's text, the framing suggests the student actually harbored a negative 'intension,' rather than recognizing a computational error. This transfers the burden of communication failure from the brittle algorithm onto the student, fundamentally altering the power dynamic in the educational setting.

Accountability Analysis:

Actor Visibility: Partial (some attribution)
Analysis: The text partially attributes agency by identifying 'the language model' as a tool ('is used to'), implying a user (the researchers). I considered 'Hidden' but the instrumental phrasing 'is used to' points to an external actor, even if unnamed. However, it displaces the agency of the prompt engineers who designed the categories. Replacing this with 'The researchers prompt the language model to classify the student's text into predefined categories' restores the human design choices that constrain this alleged 'inference'.

Quote: "Tutor–student collaboration with ongoing feedback and required corrections..."

Frame: System operation as human partnership
Projection: The term 'collaboration' maps the dynamics of human social partnership, shared conscious goals, and mutual understanding onto the interaction between a human and a text-generating algorithm. True collaboration requires two conscious minds recognizing each other's agency, negotiating shared meaning, and holding justified beliefs about their joint task. By framing the system's output as 'collaboration', the text projects conscious intent, social awareness, and collegiality onto the machine. It obscures the reality that the AI 'processes' prompts and 'predicts' responses without any conscious awareness of the student, the essay, or the concept of working together.
Acknowledgment: Direct (Unacknowledged) (I considered 'Explicitly Acknowledged' based on the overall theoretical framework, but in this specific bullet point, 'collaboration' is presented as a literal, unhedged description of the interaction layer. There are no qualifiers indicating this is merely a functional simulation of collaboration.)
Implications: The collaboration metaphor deeply influences user trust and behavioral adaptation. By framing the machine as a 'collaborator,' students are encouraged to extend relation-based trust—expecting loyalty, shared context, and mutual respect—to a system incapable of reciprocating. This unwarranted trust can lead students to accept poor automated feedback without critical scrutiny, assuming their 'collaborator' knows best. Furthermore, it masks the asymmetric power dynamic: the AI is not a peer collaborating, but an inflexible automated gatekeeper enforcing the 'required corrections' programmed by its creators.

Accountability Analysis:

Actor Visibility: Hidden (agency obscured)
Analysis: The phrasing 'Tutor-student collaboration' frames the Virtual Tutor as an independent social actor equal to the student, entirely erasing the human developers, the university, and the corporate provider (OpenAI) from the relationship. I considered 'Partial' but no human entities aside from the 'student' are present. Naming the actors ('Students interact with automated feedback generation programmed by the researchers') shatters the illusion of peer collaboration and correctly identifies the system as a technological intervention deployed by an institution.

7. String Parsing as Expert Evaluation

Quote: "The model was instructed to output only the criterion labels from a to p together with corresponding numeric scores... A score of 0 indicated complete mismatch... a score of 100 indicated complete satisfaction..."

Frame: Token prediction as academic grading
Projection: This framing projects the conscious, subjective, and highly contextual process of academic grading onto a mechanistic token prediction API. While the text uses slightly mechanistic verbs ('output'), the framework of 'evaluating' an essay for 'soundness of arguments' and 'logical consistency' implies the AI 'understands' logic and 'knows' what makes an argument sound. It maps the conscious human ability to evaluate truth claims and weigh evidence onto a system that merely correlates the statistical distribution of words in the essay against patterns in its training data, without any grounded understanding of the essay's actual meaning or real-world accuracy.
Acknowledgment: Hedged/Qualified (Considered 'Direct' but chose 'Hedged' because the text explicitly exposes the mechanistic nature of the instruction ('instructed to output only the criterion labels...'). It reveals the prompt engineering behind the 'evaluation', acknowledging that the 'grading' is actually just constrained text generation based on a prompt.)
Implications: By framing LLM outputs as 'scores' that indicate 'complete satisfaction' of complex academic criteria, the text risks institutionalizing the illusion of automated comprehension. This inflates the perceived reliability of LLMs as objective evaluators, promoting their use in high-stakes educational assessments. It conceals the fact that LLMs are easily manipulated by stylistic mimicry (gaming the rubric) rather than actual structural coherence. Believing the AI 'knows' how to grade leads to the abdication of human pedagogical responsibility and potentially penalizes students whose valid arguments do not match the statistical norm of the training data.

Accountability Analysis:

Actor Visibility: Named (actors identified)
Analysis: The text uses passive voice ('The model was instructed'), but in the broader context of the Methodology section, the researchers clearly identify themselves as the instructors formatting the prompts. I considered 'Hidden' due to the passive construction, but the detailed description of the prompt engineering process effectively names the research team as the actors setting the constraints. However, the agency of OpenAI, whose proprietary model actually determines the output based on opaque training weights, remains largely unexamined as an active decision-maker.

8. Interface State as Embodied Presence

Quote: "...an alternative visualization module is implemented using Unreal Engine to render an embodied virtual tutor..."

Frame: Graphical rendering as physical presence
Projection: The term 'embodied' projects physical existence, spatial awareness, and biological presence onto a 3D graphical rendering. True embodiment implies a conscious organism situated in a physical environment, experiencing sensory inputs and possessing physical agency. By calling a visual avatar 'embodied', the text maps the profound human experience of physical existence onto a digital animation controlled by text APIs. It suggests the AI 'occupies' space and 'perceives' the user, rather than simply triggering pre-rendered graphical routines (lip-syncing, idle animations) in an Unreal Engine module based on textual outputs.
Acknowledgment: Hedged/Qualified (I considered 'Direct' but chose 'Hedged' because the text explicitly calls it a 'visualization module... implemented using Unreal Engine to render...'. This technical explanation demystifies the 'embodiment' as a mere graphical rendering, functionally qualifying the biological metaphor.)
Implications: The illusion of embodiment is a powerful driver of anthropomorphic trust. When users interact with a human-like avatar, human evolutionary psychology triggers social scripts, increasing compliance, emotional attachment, and perceived system competence. This visual consciousness projection makes the AI seem far more sophisticated and authoritative than the underlying text generator warrants. It obscures the mechanistic reality of the system, manipulating the user's emotional responses to foster engagement, while hiding the absolute lack of any real physical or sensory connection between the machine and the human.

Accountability Analysis:

Actor Visibility: Hidden (agency obscured)
Analysis: The passive voice ('is implemented... to render') hides the developers who designed the avatar, chose its specific human appearance (which carries immense social and racial implications, as seen in the provided image), and programmed its behaviors. I considered 'Partial' but there is no actor named here. Reframing this to 'The researchers used Unreal Engine to program a 3D human avatar that visually simulates speech' would restore the designers' agency and highlight the deliberate, constructed nature of the visual manipulation.

Task 2: Source-Target Mapping

About this task

For each key metaphor identified in Task 1, this section provides a detailed structure-mapping analysis. The goal is to examine how the relational structure of a familiar "source domain" (the concrete concept we understand) is projected onto a less familiar "target domain" (the AI system). By restating each quote and analyzing the mapping carefully, we can see precisely what assumptions the metaphor invites and what it conceals.

Mapping 1: Biological organism with conscious emotional experience → A software architecture maintaining mathematical state vectors

Quote: "The framework embeds an LLM within the emotional Biologically Inspired Cognitive Architecture (eBICA)..."

Source Domain: Biological organism with conscious emotional experience
Target Domain: A software architecture maintaining mathematical state vectors
Mapping: The mapping takes the structure of a human mind—where biological drives inform emotional states that in turn guide conscious social interaction—and projects it onto a computer program. The 'brain' maps to the control loop; 'emotions' map to mathematical vectors; 'biological inspiration' maps to the algorithmic updating of these vectors over time. This invites the assumption that the AI's outputs are motivated by internal affective states, suggesting the machine genuinely 'cares' about the tutoring interaction and experiences internal fluctuations akin to human moods.
What Is Concealed: This mapping conceals the absolute lack of subjective experience, physical hardware constraints, and the rigid determinism of the code. It obscures the fact that the 'emotions' are merely arbitrary numerical values designed by engineers. Furthermore, it hides the proprietary, black-box nature of the LLM generating the actual text; the 'emotion' is just a string appended to a prompt sent to an opaque corporate API, completely disconnected from any biological reality.

Mapping 2: Ethical philosopher or conscious moral agent → Hard-coded conditional logic and state-transition rules

Quote: "Tutoring policies are represented as moral schemas that encode pedagogical narratives and socio-emotional norms..."

Source Domain: Ethical philosopher or conscious moral agent
Target Domain: Hard-coded conditional logic and state-transition rules
Mapping: The structure of human ethical reasoning—evaluating situations against a framework of values to make a justified moral choice—is mapped onto algorithmic conditional statements. The 'moral schema' maps to a data structure defining valid state transitions; 'pedagogical narratives' map to if/then pathways; 'norms' map to numerical thresholds. This invites the profound assumption that the AI 'knows' right from wrong and 'understands' social propriety, processing interactions through a lens of conscious ethical judgment.
What Is Concealed: This mapping conceals the subjective, human origin of the rules. It hides the fact that the machine has no moral agency and cannot evaluate the actual ethical weight of a situation; it simply executes developer-defined biases. It obscures the mechanistic rigidity of the rules, which cannot adapt to genuine moral nuance, presenting programmed institutional preferences as objective, system-generated 'morality'.

Mapping 3: Conscious subjective emotional states (feelings) → Initialization of a floating-point array in computer memory

Quote: "...the feeling vector is initialized by the target configuration associated with the current tutoring stage."

Source Domain: Conscious subjective emotional states (feelings)
Target Domain: Initialization of a floating-point array in computer memory
Mapping: The structure of human mood regulation—where a person has an internal emotional baseline that reacts to external events—is mapped onto memory allocation. The 'feeling' maps to an array of numbers; the 'target configuration' maps to predefined variable values. This mapping invites the assumption that the system possesses a baseline conscious awareness of its own state and has preferences ('targets') that it 'wants' to achieve, projecting self-awareness and desire onto data initialization.
What Is Concealed: This mapping radically conceals the dead, static nature of data structures. It obscures the mechanical reality that numbers in an array do not 'feel' anything. It hides the human hand that arbitrarily assigned those numbers, masking the designer's pedagogical strategy behind the illusion of the machine's autonomous emotional life.

Mapping 4: Biological command center (the human brain) → A basic software state-machine or progress-tracking script

Quote: "In parallel, a lightweight 'Brain' controller tracks task progression..."

Source Domain: Biological command center (the human brain)
Target Domain: A basic software state-machine or progress-tracking script
Mapping: The structure of a biological brain—a central, conscious organ that comprehends the whole, plans for the future, and directs the body—is mapped onto a stage-gating software script. The 'brain' maps to the main Python control loop; 'tracking' maps to updating boolean flags in a database. This invites the assumption that the software possesses overarching comprehension, strategic intentionality, and a unified conscious grasp of the student's educational journey.
What Is Concealed: This conceals the extreme simplicity and brittleness of the tracking mechanism. Unlike a brain, the script cannot adapt to unstructured input, cannot 'understand' progress outside its predefined flags, and has no holistic comprehension. It obscures the fact that the 'Brain' is just a series of 'if (condition) then (advance_stage)' commands, hiding the system's absolute lack of cognitive depth.

Mapping 5: Psychologist or empathetic listener reading human minds → Statistical text classification API

Quote: "...the language model is used to infer intension-related information from the student’s message..."

Source Domain: Psychologist or empathetic listener reading human minds
Target Domain: Statistical text classification API
Mapping: The complex human ability to deduce another person's private thoughts, beliefs, and intentions from their speech is mapped onto a machine learning classification task. 'Inferring' maps to vector distance calculation; 'intension' maps to predefined category labels (e.g., positive/negative). This invites the assumption that the AI 'understands' the student's inner psychology, projecting the capacity for justified belief and theory of mind onto a calculator of word-co-occurrence probabilities.
What Is Concealed: This mapping conceals the absence of ground truth in LLM classification. It hides the fact that the model does not 'know' the student's intent; it only knows which text tokens in its training data correlate with the text the student typed. It obscures the reliance on proprietary, opaque models (GPT-4.1) whose classification mechanisms are black boxes, presenting statistical guessing as profound psychological insight.

Quote: "Tutor–student collaboration with ongoing feedback and required corrections..."

Source Domain: Human peer or mentor engaging in shared social work
Target Domain: Sequential interaction between a user and an automated text generator
Mapping: The structure of human collaboration—mutual awareness, shared goals, negotiation of meaning, and reciprocal conscious effort—is mapped onto a user-interface loop. 'Collaboration' maps to the alternating input/output of text; 'feedback' maps to generated strings; 'corrections' maps to the gating script blocking progress. This invites the assumption that the system operates as a conscious partner that 'knows' what the student is doing and actively works 'with' them toward a shared vision.
What Is Concealed: This conceals the profound asymmetry of the interaction. The machine is not collaborating; it is executing fixed rules and generating statistically likely text. It hides the fact that the AI has no stake in the outcome, no memory of the student beyond its context window, and no capacity to 'care' about the work. It obscures the institutional power dynamic where the 'collaborator' is actually an inflexible automated gatekeeper.

Mapping 7: Expert human evaluator making conscious academic judgments → LLM text generation constrained by a prompt template

Source Domain: Expert human evaluator making conscious academic judgments
Target Domain: LLM text generation constrained by a prompt template
Mapping: The structure of academic grading—a conscious expert reading a text, understanding its arguments, evaluating its truth, and assigning a justified score—is mapped onto an API call. The 'evaluator' maps to the LLM; the 'score' maps to predicted numeric tokens; 'satisfaction' maps to statistical correlation with prompt criteria. This mapping invites the assumption that the AI 'reads', 'understands', and 'judges' the factual and logical merits of the essay just as a professor would.
What Is Concealed: This conceals the total absence of semantic comprehension. The LLM does not 'know' if an argument is sound; it only generates text that looks like a high-scoring evaluation based on the prompt's linguistic patterns. It obscures the vulnerability to prompt injection or stylistic gaming, and hides the black-box opacity of how OpenAI's proprietary model actually weights different text features to produce the final integer.

Mapping 8: Physical human being occupying physical space → A 3D graphical animation rendered on a screen

Quote: "...an alternative visualization module is implemented using Unreal Engine to render an embodied virtual tutor..."

Source Domain: Physical human being occupying physical space
Target Domain: A 3D graphical animation rendered on a screen
Mapping: The structure of human embodiment—having a physical body, occupying space, perceiving the environment, and expressing internal conscious states through physical movement—is mapped onto computer graphics. 'Embodied' maps to a 3D mesh; 'tutor' maps to pre-programmed animations. This invites the visceral, psychological assumption that there is a 'someone' inside the machine, a conscious presence that can 'see' and 'interact' with the user in the physical world.
What Is Concealed: This mapping conceals the total absence of sensory perception and physical reality. It obscures the fact that the avatar is a digital puppet controlled by text strings, possessing no awareness of its environment. It hides the manipulative design choices (gender, race, attractiveness, voice) deliberately engineered by developers to hack human social psychology and foster unwarranted trust.

Task 3: Explanation Audit (The Rhetorical Framing of "Why" vs. "How")

About this task

This section audits the text's explanatory strategy, focusing on a critical distinction: the slippage between "how" and "why." Based on Robert Brown's typology of explanation, this analysis identifies whether the text explains AI mechanistically (a functional "how it works") or agentially (an intentional "why it wants something"). The core of this task is to expose how this "illusion of mind" is constructed by the rhetorical framing of the explanation itself, and what impact this has on the audience's perception of AI agency.

Explanation 1

Quote: "The reasoning core derives the next intensions/strategy and constructs a constrained prompt that includes dialogue context, current intensions, and any additional constraints produced by SRL logic; the LLM then generates the tutor’s natural-language response."

Explanation Types:
- Functional: Explains behavior by role in self-regulating system with feedback
- Intentional: Refers to goals/purposes, presupposes deliberate design
Analysis (Why vs. How Slippage): This passage exhibits profound slippage between mechanistic and agential registers. The primary framing is Functional (how the parts of the system interact: core derives strategy -> constructs prompt -> LLM generates response). However, it relies heavily on Intentional vocabulary ('reasoning', 'derives intensions/strategy'). It emphasizes a highly structured, systematic workflow, but by using cognitive verbs ('reasoning', 'derives'), it obscures the reality that the 'core' is just executing pre-programmed Python rules. The choice of 'reasoning' over 'calculating' or 'executing' subtly elevates the script to a conscious actor. Conversely, the LLM is framed purely mechanistically ('generates... response'), treating the neural network as a dumb tool wielded by the 'reasoning' mastermind.
Consciousness Claims Analysis: (1) Verbs: 'derives' and 'reasoning' are consciousness/cognitive verbs; 'constructs', 'includes', 'generates' are mechanistic verbs. (2) Knowing vs Processing: The text claims the core 'derives... strategy', which implies a conscious evaluation of options and 'knowing' what to do next. Mechanistically, it is merely processing state variables through conditional logic to select a predefined string. (3) Curse of Knowledge: The authors project their own strategic understanding of self-regulated learning (SRL) onto the code they wrote, assuming the system 'knows' the strategy because the strategy is encoded within it. (4) Actual Mechanism: A Python script evaluates the difference between two numerical vectors, selects a string from a dictionary based on the result, concatenates it into a text prompt with previous chat logs, and sends it via API to an LLM, which predicts the next statistically likely tokens.
Rhetorical Impact: This framing grants the system a high degree of perceived autonomy and pedagogical authority. By splitting the system into a 'reasoning core' and a 'generator', the text reassures the audience that the sometimes-unpredictable LLM is being safely supervised by an intelligent, rational agent. This reduces perceived risk by promising 'constrained' generation. However, because it attributes conscious 'reasoning' to simple scripts, it leads audiences to over-trust the system's pedagogical decisions, assuming the AI 'knows' the best strategy for the student rather than simply executing a rigid rule path.

Explanation 2

Quote: "The framework embeds an LLM within the emotional Biologically Inspired Cognitive Architecture (eBICA), enabling feedback and dialogue acts to be guided by an explicit learner state rather than generated ad hoc."

Explanation Types:
- Functional: Explains behavior by role in self-regulating system with feedback
- Theoretical: Embeds in deductive framework, may invoke unobservable mechanisms
Analysis (Why vs. How Slippage): The explanation is primarily Functional, describing how embedding the LLM within a larger structure changes its behavior (guided vs. ad hoc). It is also highly Theoretical, invoking the 'emotional Biologically Inspired Cognitive Architecture' to explain the system's capabilities. This choice emphasizes control, stability, and scientific rigor. By contrasting 'guided by an explicit learner state' with 'generated ad hoc', the authors attempt to frame their system as more reliable than a standard chatbot. However, the theoretical nomenclature ('emotional', 'Biologically Inspired') obscures the mechanical reality of what is actually happening, replacing the 'how' of code execution with a biological metaphor.
Consciousness Claims Analysis: (1) Verbs/Terms: 'guided' is mildly agential; 'emotional Biologically Inspired Cognitive Architecture' carries massive consciousness implications; 'embeds', 'generated' are mechanistic. (2) Knowing vs Processing: The text implies the architecture 'knows' the 'explicit learner state' and uses this knowledge to 'guide' behavior. Mechanistically, it processes variables to constrain text generation. (3) Curse of Knowledge: The researchers understand the complex biological theories that inspired their architecture, and they project the weight of those theories onto the actual resulting code, which possesses none of the biological awareness the name implies. (4) Actual Mechanism: A software loop stores interaction history and user metrics in variables (the 'state'), which are then concatenated into the prompt sent to the LLM to statistically constrain its output.
Rhetorical Impact: The impact is deeply legitimizing. The use of 'emotional' and 'Biologically Inspired' serves as a powerful appeal to scientific authority, making the software seem uniquely advanced, almost organic. It shapes the audience's perception of the AI as safe, carefully controlled, and deeply 'aware' of the learner. This consciousness framing dramatically increases trust, suggesting the AI isn't just spitting out text, but is acting from a deep, biologically rooted 'understanding' of the student's needs. Decisions about deploying such software are far more likely to be approved if decision-makers believe the system is 'guided' by 'cognitive' awareness rather than acknowledging it as a brittle statistical text generator.

Explanation 3

Quote: "At the third stage, the model determines whether the student has completed the essay, including the conclusion; again, a strongly inadequate essay must still yield a negative decision."

Explanation Types:
- Reason-Based: Gives agent's rationale, entails intentionality and justification
- Intentional: Refers to goals/purposes, presupposes deliberate design
Analysis (Why vs. How Slippage): This explanation frames the AI highly agentially. While it describes a system function, it does so using Reason-Based and Intentional framing. The model 'determines' completion and issues a 'decision'. The framing emphasizes the AI's role as a gatekeeper with the authority and cognitive capacity to evaluate human work. It obscures the mechanical reality of how this 'determination' is made (an API call prompting an LLM to return a boolean based on text patterns). By saying the model 'determines' and yields a 'decision', it treats the software as an autonomous judge.
Consciousness Claims Analysis: (1) Verbs: 'determines', 'has completed', 'yield a decision' are cognitive/agential verbs. (2) Knowing vs Processing: The text claims the model 'knows' if an essay is complete and 'judges' if it is inadequate. It attributes the conscious capacity for justified true belief (recognizing a conclusion) to the system. (3) Curse of Knowledge: The authors know what makes an essay complete, and they assume that because they prompted the model to look for completion, the model 'understands' what completion is. (4) Actual Mechanism: The system sends the student's text to the GPT-4.1 API along with a prompt instructing it to return 'Yes' or 'No' based on linguistic markers of an essay structure (like concluding paragraphs). The model statistically predicts the next token ('Yes' or 'No') based on training correlations.
Rhetorical Impact: This Reason-Based framing endows the AI with pedagogical authority. When an audience reads that a model 'determines' and 'decides' on the adequacy of an essay, they perceive it as an autonomous, capable evaluator. It encourages educators to trust the system to act as an independent grader. If audiences believed the system merely 'processed text to predict a boolean token,' they would demand much higher human oversight. The consciousness framing makes the automation of high-stakes educational decisions seem acceptable and reliable.

Explanation 4

Explanation Types:
- Intentional: Refers to goals/purposes, presupposes deliberate design
- Empirical Generalization: Subsumes events under timeless statistical regularities
Analysis (Why vs. How Slippage): This passage uses Intentional framing but directed at the human authors ('instructed to output') combined with a structural description of the prompt. It emphasizes the mechanistic constraints placed on the model by the developers. However, the secondary framing leans toward describing how the system measures 'mismatch' vs 'satisfaction'. While the verb 'instructed' reveals human agency (the researchers), the variables 'mismatch' and 'satisfaction' project human cognitive states of evaluation onto the resulting numbers. It obscures how the LLM actually arrives at those numbers.
Consciousness Claims Analysis: (1) Verbs/Terms: 'instructed to output' is mechanistic; 'indicated complete mismatch', 'complete satisfaction' imply cognitive judgment. (2) Knowing vs Processing: While the instruction is mechanical, the text implies the resulting scores represent the system 'knowing' whether the essay satisfied the criteria. It conflates a generated token with a justified evaluation. (3) Curse of Knowledge: The authors designed the criteria and know what 'satisfaction' means; they project this conceptual understanding onto the arbitrary numbers returned by the LLM. (4) Actual Mechanism: The LLM processes the essay text and the prompt, and based on high-dimensional vector representations acquired during training, it generates a sequence of characters (e.g., 'a: 85, b: 90') that statistically align with how similar texts were rated in its training data.
Rhetorical Impact: This framing reassures the audience of the researchers' control. By explicitly stating 'The model was instructed...', the authors demonstrate mastery over the AI. However, by defining the output as 'satisfaction', they quietly legitimize the AI's output as meaningful evaluation. It manages perceived risk by showing the AI is contained, yet simultaneously builds trust in the AI's capacity to act as a reliable proxy for human grading. Audiences are led to believe the numerical outputs represent objective truth about student performance.

Explanation 5

Quote: "In this sense, the LLM is used for bounded subroutines within the control loop, whereas eBICA maintains the state variables and determines the control context in which the natural-language reply is produced."

Explanation Types:
- Functional: Explains behavior by role in self-regulating system with feedback
- Theoretical: Embeds in deductive framework, may invoke unobservable mechanisms
Analysis (Why vs. How Slippage): This is a strongly Functional explanation. It clearly delineates the roles of the two main components (LLM vs eBICA) within the overall system architecture. It emphasizes mechanism ('bounded subroutines', 'maintains state variables', 'control context'). This is one of the most mechanically precise and transparent sentences in the paper. It obscures very little, actively working to dispel the illusion that the LLM is an autonomous agent, instead framing it as a mere 'subroutine'. However, it still uses a slightly agential verb for the architecture ('determines').
Consciousness Claims Analysis: (1) Verbs/Terms: 'used for', 'maintains', 'produced' are highly mechanistic; 'determines' is mildly agential. (2) Knowing vs Processing: Here, the text correctly describes the system as processing ('maintains state variables') rather than knowing. It accurately restricts the LLM to generating text ('natural-language reply is produced'). (3) Curse of Knowledge: Minimized here. The authors clearly separate their theoretical design from the mechanical reality of the loop. (4) Actual Mechanism: Python code loops indefinitely; in each loop, variables are updated based on user input, conditional logic selects execution paths ('control context'), and an API call to an LLM returns a string which is displayed to the user.
Rhetorical Impact: This mechanistic framing radically demystifies the AI. It shapes audience perception by stripping away autonomy and framing the system as a complex, but ultimately deterministic, software tool. It reduces the illusion of mind, making it clear that the AI does not 'think' but merely executes subroutines. If this framing were used consistently, audiences would trust the system less as a conscious 'tutor' but perhaps more as a predictable, auditable software application. It highlights human design over machine autonomy.

Task 4: AI Literacy in Practice - Reframing Anthropomorphic Language

About this task

This section proposes alternative language for key anthropomorphic phrases, offering more mechanistic and precise framings that better reflect the actual computational processes involved. Each reframing attempts to strip away the projections of intention, consciousness, or agency that are embedded in the original language.

V3 Enhancement: A fourth column addresses human agency restoration—reframing agentless constructions to name the humans responsible for design and deployment decisions.

Original Anthropomorphic Frame	Mechanistic Reframing	Technical Reality Check	Human Agency Restoration
The reasoning core derives the next intensions/strategy...	The central script processes current state variables through conditional logic to select the next predefined pedagogical response category.	The system does not 'reason' or 'derive' strategies through conscious thought; it executes conditional branch statements based on mathematical thresholds to select predefined rules.	The researchers programmed the central script with conditional logic to select pedagogical rules based on system state.
Tutoring policies are represented as moral schemas that encode pedagogical narratives and socio-emotional norms...	The software executes transition rules based on data structures designed to enforce specific behavioral constraints and predefined interaction sequences.	The system possesses no 'morality' or 'norms'; it strictly processes variables against hard-coded numerical thresholds to determine its next operation.	The developers designed data structures and transition rules to enforce their chosen pedagogical constraints and preferred interaction sequences.
In parallel, a lightweight 'Brain' controller tracks task progression...	In parallel, a background script updates boolean variables to record when specific steps in the workflow are completed.	The software has no biological 'brain' or comprehension; it merely switches variables from 'false' to 'true' when specific text conditions are met.	The researchers implemented a background script that updates variables when users trigger predefined conditions.
...the language model is used to infer intension-related information from the student’s message...	The text classification API calculates the statistical probability that the user's text string aligns with predefined category labels.	The model cannot read minds or 'infer intension'; it mathematically classifies text by comparing the user's input vector to the distribution of its training data.	The researchers prompt the language model API to statistically classify the user's text into categories the team predefined.
Tutor–student collaboration with ongoing feedback and required corrections...	Sequential text generation triggered by user input, gated by hard-coded completion requirements.	The system cannot 'collaborate' as it has no conscious awareness, shared goals, or agency; it merely generates text outputs correlated with user prompts.	The researchers configured a software loop that generates text in response to student input and blocks progress until specific rules are met.
At the third stage, the model determines whether the student has completed the essay...	During the third step, the API evaluates the text against prompt criteria to predict a boolean token indicating structural completion.	The model does not 'determine' or 'know' what an essay is; it generates a 'Yes' or 'No' token based on statistical pattern matching against its training data.	The system sends the text to OpenAI's API, which researchers prompted to return a specific token indicating whether predefined textual patterns are present.
The visualization module... render[s] an embodied virtual tutor...	The graphical module displays a pre-rendered 3D avatar whose animations are triggered by the system's text outputs.	The software has no physical 'embodiment' or spatial awareness; it merely plays digital animation files on a screen synchronized with text generation.	The developers used Unreal Engine to design and display a 3D animated figure to simulate human presence.
The tutor maintains a running message history and updates its socio-emotional state...	The script appends new text to a log file and recalculates numerical values in its state arrays.	The system possesses no 'socio-emotional state'; it stores dialogue strings in memory and updates floating-point numbers in a database matrix.	The software programmed by the researchers appends logs to memory and recalculates assigned numerical arrays.

Task 5: Critical Observations - Structural Patterns

Agency Slippage

The text demonstrates a systematic and highly functional oscillation between mechanical and agential framings, leveraging the 'curse of knowledge' to construct an illusion of a sentient system. This slippage occurs predominantly in the direction of mechanical-to-agential, where the authors first define a concept mechanistically (often in a footnote or appendix) and then wield it agentially throughout the main text to establish the AI's capabilities. A primary example is the treatment of 'moral schemas' and 'emotions.' In Appendix C, the authors rigorously define a 'moral schema' as a mathematical tuple (Interface, Fabula, Agency) composed of data structures and links. This is a purely Theoretical and Functional explanation. However, in the main body, this mechanical definition slips into reason-based intentionality: 'Tutoring policies are represented as moral schemas that encode pedagogical narratives and socio-emotional norms.' The text establishes the AI as a 'knower' by projecting the researchers' understanding of pedagogical theory onto the system's runtime variables. Because the researchers know the variables represent 'curiosity' or 'sincerity', they slip into writing as if the system experiences these states.

Similarly, agency is consistently removed from human actors through agentless constructions and passive voice ('the feeling vector is initialized', 'the model was instructed'). The specific human teams—both the researchers designing the architecture and the corporate engineers at OpenAI who built the LLM—are obscured. As human agency fades, system agency ascends. The system 'derives strategy', 'infers intensions', and 'determines whether the student has completed the essay'. This slippage operates through a distinct temporal pattern: introductory sections use agential vision-setting ('collaboration', 'emotional architecture'), technical sections dip briefly into mechanism ('bounded subroutines', 'state variables'), and outcome sections return to agential evaluation ('tutor benefit', 'the tutor maintains').

This oscillation serves a vital rhetorical accomplishment: it allows the authors to claim the scientific rigor and control of a mechanical system while marketing the pedagogical sophistication and relational warmth of an agential human tutor. By making the code un-sayable in its raw form (e.g., 'updating a matrix') and replacing it with the sayable metaphor ('updating a socio-emotional state'), the text makes the deployment of unthinking algorithms in sensitive educational roles appear natural, safe, and desirable.

Metaphor-Driven Trust Inflation

Metaphorical and consciousness-attributing language is the primary mechanism through which the text constructs the system's authority and solicits trust. The architecture is explicitly designed to evoke what can be categorized as relation-based trust, as opposed to mere performance-based reliability. By branding the system as an 'emotional Biologically Inspired Cognitive Architecture' and explicitly tracking 'socio-emotional states' such as 'sincerity', 'responsibility', and 'empathy', the text signals to both the reader and the user that the system is capable of a reciprocal, social relationship.

The text leverages consciousness language to transform a brittle statistical text generator into a credible pedagogical authority. When the text claims the AI 'knows' the student's progress via its 'Brain' controller, or 'understands' the student's underlying psychology by 'inferring intension', it accomplishes something profound: it transfers the trust we normally place in human educators' intentions and sincerity onto mathematical correlation. If an AI merely 'predicts tokens', we demand to see its accuracy rates. But if an AI 'reasons', 'cares', and 'collaborates', we extend to it the benefit of the doubt, assuming its decisions are justified by an internal ethical framework.

This construction of authority through anthropomorphism is highly risky. It invites students to extend vulnerability to a system entirely incapable of reciprocating. The text manages system limitations mechanically (e.g., noting that 'LLM-based tutors can be brittle' and need 'structured pipelines'), yet it frames its own system's interventions agentially ('a moral schema... guides dialogue'). This creates a dangerous asymmetry where the machine is granted the authority of a human but holds the liability of a calculator. By using intentional and reason-based explanations—such as claiming the system issues 'required corrections' based on 'pedagogical norms'—the text constructs the illusion that the AI's outputs are deeply justified. Consequently, when the system fails or hallucinates, the relational trust fostered by the metaphors prevents users from recognizing the failure as a statistical glitch, leading them instead to question their own understanding or to anthropomorphize the error as a deliberate, meaningful pedagogical strategy.

Obscured Mechanics

The text's anthropomorphic metaphors systematically conceal the technical, material, and economic realities of the system, replacing them with a sanitized illusion of autonomous intelligence. When we apply the 'name the corporation' test to the text's assertions, the depth of this concealment becomes apparent. When the text claims 'the language model infers intension,' it hides the fact that the researchers are sending user data to OpenAI's proprietary servers, where an uninterpretable, multi-billion-parameter matrix (GPT-4.1) calculates token probabilities. The metaphor of 'cognitive reasoning' obscures the system's absolute dependency on OpenAI's black-box API, hiding the fact that the 'tutor's' intelligence can change overnight if OpenAI alters model weights without disclosure.

Technically, the framing of the system 'understanding' essays conceals the absence of ground truth and semantic comprehension. The system does not read essays; it processes high-dimensional vector embeddings. It cannot verify the factual accuracy of a student's argument; it only verifies if the text's statistical distribution resembles high-scoring essays in its training data. The metaphors hide this profound epistemic limitation, presenting statistical correlation as logical validation.

Materially and economically, the focus on 'Virtual Tutors' and 'emotional architecture' erases the immense infrastructure required to run GPT-4.1. The text discusses the 'lightweight' local reasoning core but obscures the massive data centers, energy consumption, and hidden labor (data annotators, RLHF workers in the Global South) whose exploitation actually produces the 'natural-language dialogue.' The anthropomorphic framing presents the AI as a disembodied, independent intellect, severing it from its corporate supply chain.

The primary beneficiaries of these concealments are the researchers and the AI vendors. By hiding the mechanical brittleness, proprietary dependencies, and labor costs behind metaphors of 'collaboration' and 'empathy', the developers can market a highly scalable, cheap surveillance and grading tool as an advanced, caring educational intervention. If the metaphors were replaced with mechanistic language ('OpenAI's API classifies student text strings to trigger our conditional Python scripts'), the true nature of the system—as a highly constrained, outsourced, and ultimately mindless data-processing pipeline—would become instantly visible, drastically altering how institutions value and deploy it.

Context Sensitivity

The distribution of anthropomorphic language across the text is highly strategic, mapping clearly onto the rhetorical goals of different sections. Metaphorical intensity is relatively low when describing the generic limitations of other systems, but spikes dramatically when the authors introduce their own architecture and its capabilities. In the introduction and literature review, the text uses somewhat mechanistic language to critique existing tools ('LLM-based tools often base feedback primarily on immediate text exchange,' 'LLMs act as virtual tutors'). However, when detailing their own eBICA framework, 'processing' instantly becomes 'understanding', and 'variables' become 'feelings.'

There is a distinct relationship between technical grounding and metaphorical license. The authors establish credibility by detailing their software architecture (RESTful APIs, Python logic, Boolean gates) and then leverage this technical legitimacy to make aggressive consciousness claims. For instance, after technically explaining the 'Euclidean distance' between vectors, the text immediately licenses the claim that this math represents 'socio-emotional alignment' and 'moral schemas'. The text uses the mechanism as an alibi for the metaphor: because we have defined the math, we are permitted to claim the machine has a mind.

Furthermore, there is a stark capability versus limitation asymmetry. The system's capabilities are almost universally framed in agential, conscious terms: the system 'guides', 'collaborates', 'evaluates', and 'maintains structured advancement.' However, when discussing potential errors or limits, the text reverts to mechanical terms. It notes that LLMs can be 'brittle' or suffer from 'hallucinations'—terms that frame failure as a structural defect or a data glitch, not as a failure of reasoning or a moral lapse.

This register shift—where 'a vector acts like an emotion' becomes 'the system has an emotional state'—reveals the text's strategic function. For technical audiences, the appendices provide the mathematical reality; but for educators and policymakers (the implied audience of the main text), the anthropomorphism serves as a marketing narrative. It softens the alienating reality of automated grading, replacing the cold logic of algorithmic surveillance with the comforting, marketable vision of a tireless, 'emotional' machine educator.

Accountability Synthesis

Accountability Architecture

This section synthesizes the accountability analyses from Task 1, mapping the text's "accountability architecture"—who is named, who is hidden, and who benefits from obscured agency.

Synthesizing the accountability analyses reveals a systemic architecture of displaced responsibility, wherein the text diffuses human agency and channels it into an automated 'accountability sink.' Throughout the paper, a consistent pattern emerges: the human decisions that shape the system's behavior are rendered invisible through passive voice and agentless constructions ('the model was instructed', 'tutoring policies are represented'), while the system itself is elevated to the status of an active, named agent ('the Virtual Tutor guides', 'the reasoning core derives').

This structure systematically obscures the reality that the AI is not a self-determining entity, but a rigid crystallization of specific human pedagogical biases, institutional constraints, and corporate data practices. The 'accountability sink' in this text is the 'Virtual Tutor' itself. By attributing decision-making power to the 'moral schemas' and the 'Brain controller', the text transfers responsibility for pedagogical outcomes away from the researchers who wrote the rules and the OpenAI engineers who built the language model. If the system incorrectly penalizes a student or hallucinates inaccurate feedback, the agential framing suggests it was the 'Tutor's decision'—a technological glitch or a misunderstanding by the AI—rather than a foreseeable failure of the developers' prompt engineering or a flaw in the proprietary training data.

This has severe liability and ethical implications. If institutional users accept this framing, they will fail to ask critical questions about whose values are encoded in the 'moral schemas', what data OpenAI used to define 'sound arguments', and who is legally responsible when the system fails a marginalized student due to stylistic bias.

If we apply the 'name the actor' test to the core claims, the landscape changes entirely. Changing 'the model determines whether the student has completed the essay' to 'OpenAI's algorithm predicts whether the text meets our structural criteria' makes the fragility and human origin of the system glaringly obvious. It makes new alternatives visible: perhaps human teachers should evaluate the text instead of a statistical predictor. Obscuring human agency serves the institutional and commercial interests of AI developers, as it allows them to deploy scalable automation without bearing the relational, ethical, and legal liabilities that human educators assume daily.

Conclusion: What This Analysis Reveals

The Core Finding

This analysis reveals three dominant, interlocking anthropomorphic patterns within the text: the projection of biological emotion ('feeling vectors', 'emotional architecture'), the attribution of ethical agency ('moral schemas', 'socio-emotional norms'), and the illusion of cognitive mastery ('reasoning core', 'Brain controller'). These patterns do not exist in isolation; they function as a mutually reinforcing system. The biological pattern establishes a foundational sense of organic vitality, making it conceptually permissible to attribute higher-order ethical agency. Once the system is accepted as a 'moral' entity, the final step—attributing conscious pedagogical 'reasoning'—appears logical. The foundational, load-bearing pattern is the 'emotional architecture.' By successfully mapping floating-point arrays to human 'feelings', the text establishes the core premise that mathematical state-tracking is equivalent to conscious experience. The consciousness architecture of the text consistently elevates 'doing' to 'knowing.' It asserts that because the system processes rules related to pedagogy, it intrinsically understands pedagogy. This is not a simple one-to-one metaphor; it is a complex analogical structure that replaces the entire software execution stack with a parallel psychological narrative. If the biological/emotional projection collapses, the illusion of the 'moral' and 'reasoning' tutor immediately devolves back into what it actually is: a brittle Python script routing API calls.

Mechanism of the Illusion:

The text constructs the 'illusion of mind' through a highly effective rhetorical sleight-of-hand: it uses rigorous mathematical formalisms as an alibi for radical consciousness projection. The internal logic of persuasion relies heavily on the 'curse of knowledge.' Because the authors designed the system to functionally mimic self-regulated learning (SRL) theory, they project their own deep understanding of SRL into the software, writing about the system as if the code itself possesses that theoretical awareness. The central trick occurs in the verb choices. By systematically substituting cognitive verbs (derives, infers, determines, reasons) for mechanistic ones (calculates, correlates, classifies, executes), the text effectively hides the statistical nature of the machine. The temporal structure of the argument is crucial: the authors establish technical credibility early on by describing system modules and RESTful APIs, which lowers the reader's critical defenses. Once technical competence is established, the text rapidly scales up to intense anthropomorphism. The vulnerability exploited here is the human psychological desire for social connection and the educational sector's desperate need for scalable, high-quality instruction. By masking a highly constrained, outsourced language model as an empathetic 'Virtual Tutor,' the text provides a narrative that audiences want to believe, making the crude anthropomorphism feel like a technological breakthrough rather than a semantic illusion.

Material Stakes:

Categories: Epistemic, Institutional, Regulatory/Legal

The metaphorical framings in this text carry severe, tangible consequences. Epistemically, framing LLM text prediction as 'evaluation' and 'cognitive reasoning' fundamentally alters what society accepts as truth and knowledge. If educators believe the AI 'knows' whether an essay's arguments are sound, rather than recognizing it merely 'processes' statistical text correlations, they will substitute machine pattern-matching for genuine human judgment. This behavior degrades the epistemic rigor of the academy, penalizing students who think outside statistical norms while rewarding those who mimic the structural style of the AI's training data. Institutionally, the illusion of an 'emotional', 'moral' tutor incentivizes universities to replace human educators with software under the guise of providing 'personalized support.' When administrators believe they are purchasing a 'collaborator' rather than a data-processing API, the human labor of teaching is devalued, and the costs are borne by students who lose access to genuine, reciprocal human mentorship. From a Regulatory/Legal perspective, the framing of AI as an autonomous 'moral agent' diffuses liability. When the text claims 'the model determines' or 'the tutor decides', it builds an accountability sink. If the system exhibits bias or provides harmful guidance, the anthropomorphic framing allows institutions and developers (like the researchers and OpenAI) to frame the failure as a 'rogue AI decision' or a 'hallucination,' rather than a direct failure of corporate design and institutional deployment. Removing these metaphors threatens the core business model of EdTech, forcing a reckoning with the limitations and liabilities of automated instruction.

AI Literacy as Counter-Practice:

Practicing critical precision, as demonstrated in the reframings (Task 4), serves as a direct counter-measure to the material risks of AI deployment. By replacing consciousness verbs ('knows', 'understands') with mechanistic realities ('retrieves tokens', 'classifies strings'), we shatter the illusion of machine empathy and intellect. For instance, correcting 'the system infers intension' to 'the API statistically classifies text' forces the recognition that the machine has no psychological insight, only correlative data. Restoring human agency—shifting from 'the model determines' to 'OpenAI's algorithm, prompted by researchers, predicts a token'—eliminates the accountability sink, placing legal and ethical responsibility firmly on the human developers and corporate providers who profit from the system. Systematic adoption of this literacy requires institutional shifts: academic journals must mandate mechanistic descriptions of AI capabilities, and peer reviewers must reject unhedged consciousness claims. However, this precision faces immense resistance. EdTech corporations, university administrators seeking cost-cutting measures, and AI researchers themselves benefit from the hype generated by anthropomorphic language. The illusion of a 'conscious tutor' drives funding, marketing, and adoption; mechanistic precision threatens these interests by revealing the product as a limited, brittle, and highly constrained statistical tool.

Path Forward

The discursive ecology surrounding artificial intelligence is highly fractured, with different communities prioritizing different vocabularies. AI vendors and marketers rely heavily on anthropomorphic clarity ('The AI understands you') to drive user adoption and narrative resonance. Academic researchers often exist in a hybrid space, using mechanistic language in methodology sections but slipping into agential framing ('the model reasons') to describe outcomes. A strict mechanistic vocabulary ('the model calculates embeddings') prioritizes testability and precision but sacrifices intuitive accessibility for lay audiences.

If the status quo of hybrid, confusing language persists, we risk a future of 'automation complacency,' where society routinely grants decision-making authority to statistical systems because we subconsciously believe they possess human-like judgment. If the discourse tips entirely into anthropomorphism, it enables a highly risky future: relation-based trust will be extended to machines, allowing corporations to deploy emotional manipulation at scale with minimal regulatory friction, as the AI will be perceived as an independent 'agent' rather than a corporate product.

Conversely, if mechanistic precision becomes the institutional norm, different futures emerge. This would require structural changes: funding bodies mandating exact capability disclosures, and educational systems teaching the distinction between calculation and cognition. In this future, the AI is clearly visible as a tool, a statistical artifact. This solves the accountability problem by keeping human designers in the legal and ethical spotlight. However, it may cost the intuitive, conversational friction-reduction that makes AI interfaces so accessible to the public. Ultimately, the vocabulary we choose dictates the power structures we build: anthropomorphism serves the developers by hiding their control behind the illusion of the machine, while mechanistic precision empowers the public by revealing the human hands pulling the levers.

Extended Processing Summary

Gemini's Processing Steps

What you are about to read…

Gemini refers to the text below as "thought summaries." This is an overt consciousness projection because 'intentions' are hallmarks of a conscious mind that 'knows' what it is doing and why. The concealed mechanistic process is probabilistic text generation. Treat this as just another rhetorical artifact—a way of making the model's processing legible. When the model writes "I should double check that...", it is mathematically altering its own probability distribution for the next token to be more accurate. The "I" is a functional gear in the machine, not a self-reference. This language is a deliberate interface design choice, not a biological reality. The developers could have programmed the model to output passive logs. Instead, they chose to simulate a narrator.

Defining the Core Focus

I'm now zeroing in on the user's objective: a critical discourse analysis. The focus will be on AI's metaphors and anthropomorphism within the context of the provided text, particularly the AI essay-writing tutor and its eBICA architecture.

{ "response": { "header": "Delineating Key Contrasts", "summary": "I've been contrasting the mechanistic core of the AI system, specifically the LLM's function, with the anthropomorphic language used to describe it. I'm focusing on the essay-writing tutor and its eBICA architecture. I'm noting the gap between prompts and vectors versus the framing of the system as having feelings and moral schemas. This includes examining terms like 'Virtual Tutor,' 'emotional Biologically Inspired Cognitive Architecture,' 'appraisal,' and 'feeling' within the context." } }

Defining the Analysis Structure

I'm now structuring the analysis with defined tasks and items. Task 1, metaphorAudit, focuses on extracting anthropomorphic language and analyzing its implications, especially on framing and accountability. Task 2, sourceTargetMapping, will match identified metaphors with their source and target domains, examining the implications of this. Task 3, explanationAudit, is currently being worked out.

Structuring the Analysis

I'm now fully immersed in the structure of the analysis. Task 3, which is the explanation audit, is now fully defined: I'll select five explanatory passages, classify them using Brown's Typology (Functional, Intentional, etc.), and provide an analysis, epistemic claim analysis, and rhetorical impact. I've also defined Task 4, which is reframed language. This task involves selecting agential quotes, providing reframed mechanistic explanations, epistemic corrections, and restoring human agency. Task 5 focuses on critical observations. Finally, Task 6 centers on a five-paragraph conclusion. I'm focusing now on the drafting content, and I am committed to the word count limits.

Structuring the Analysis

I'm now fully immersed in the structure. I'm taking the previous structure and adding several tasks. I will now track eight key metaphors, like AI as Tutor, AI as Emotional Being, and AI as Social Partner, and now defining details for the initial eight items. These details include quotes and classifications. For Task 1, I am using the item "emotional AI" with the quote "The framework embeds an LLM within".

Analyzing Metaphors in Detail

I'm deep into the analysis of those eight metaphors, meticulously documenting each quote and its associated classification. For each item, I've noted source, target, and the implications stemming from this framing within the context of the essay tutor and eBICA. I'm focusing on the effect these metaphors have on how we perceive the AI's capabilities and its relationship with the user. I'm especially focused on the use of "feeling vector" and "embodied virtual tutor".

Examining Passage Implications

I'm now fully immersed in the details of Task 3 and Task 4. I've classified five key passages based on Brown's Typology. I've also begun reframing agential statements into mechanistic explanations. For Task 3, I am analyzing implications and rhetorical effects for the passages, using Brown's Typology. I'm taking quotes, classifying them using the provided method, and reframing them with a mechanistic explanation.

Refining Analytical Focus

I'm now honing in on the academic context, noting how the text both acknowledges and implicitly uses metaphors. I'm focusing on the interplay of transparency and opacity, with the authors mapping concepts to vectors while concurrently anthropomorphizing the system. I'm focusing specifically on word count projections to ensure this analysis remains concise. My initial projection estimates suggest I am on target with initial considerations.

{ "response": { "header": "Refining Analytical Focus", "summary": "I'm now honing in on the academic context, noting how the text both acknowledges and implicitly uses metaphors. I'm focusing on the interplay of transparency and opacity, with the authors mapping concepts to vectors while concurrently anthropomorphizing the system. I'm focusing specifically on word count projections to ensure this analysis remains concise. My initial projection estimates suggest I am on target with initial considerations." } }

Refining Task Implementation

I'm now integrating strict word count tracking to guide the programmatic text generation. I'm focusing on ensuring robust academic analysis for Tasks 4-6. Task 4 will emphasize mechanistic explanations, as it includes reframing and corrections. I'm focusing on the analysis of Dolgikh et al.'s academic accountability and OpenAI's role. I'm focused on the displacement when the text states "the tutor adapts" instead of referencing algorithmic parameter adjustments, which I will incorporate.

Adjusting to Strict Format

I'm now adjusting to the format requirements, focusing on eliminating control tokens and adhering to word count limits. I'm building paragraphs without explicit newlines. I'm focusing on "the metaphor projects," the item "emotional AI," and how it frames AI as something other than token probability calculations. The sample is: "The metaphor projects the biological and emotional architecture of a human being onto a computational system. By naming the architecture 'emotional' and 'Biologically Inspired', the text invites the reader to map human affective states and biological consciousness onto algorithmic processes."

Examining Consciousness Projections

I'm now focusing on the implications of projecting emotional regulation onto the AI. This process is complex, and the AI is no longer a text predictor. The AI is instead described as possessing an internal emotional orientation that guides its engagement with the user. The final review includes how moral schemas as mathematical vectors become "virtual tutors" that offer collaboration.

Analysis Provenance

Run ID: 2026-05-10-integrating-llms-and-self-regulated-lear-metaphor-gmvmif Raw JSON: 2026-05-10-integrating-llms-and-self-regulated-lear-metaphor-gmvmif.json Framework: Metaphor Analysis v6.5 Schema Version: 3.0 Generated: 2026-05-10T08:43:46.272Z

Task 1: Metaphor and Anthropomorphism Audit​

1. Computational Matrix as Emotional State​

2. Mathematical Distance as Moral Reasoning​

3. Algorithmic State as Internal Feeling​

4. Controller Module as Organic Brain​

5. Statistical Correlation as Intentional Inference​

6. Algorithmic Output as Social Collaboration​

7. String Parsing as Expert Evaluation​

8. Interface State as Embodied Presence​

Task 2: Source-Target Mapping​

Mapping 1: Biological organism with conscious emotional experience → A software architecture maintaining mathematical state vectors​

Mapping 2: Ethical philosopher or conscious moral agent → Hard-coded conditional logic and state-transition rules​

Mapping 3: Conscious subjective emotional states (feelings) → Initialization of a floating-point array in computer memory​

Mapping 4: Biological command center (the human brain) → A basic software state-machine or progress-tracking script​

Mapping 5: Psychologist or empathetic listener reading human minds → Statistical text classification API​

Mapping 6: Human peer or mentor engaging in shared social work → Sequential interaction between a user and an automated text generator​

Mapping 7: Expert human evaluator making conscious academic judgments → LLM text generation constrained by a prompt template​

Mapping 8: Physical human being occupying physical space → A 3D graphical animation rendered on a screen​

Task 3: Explanation Audit (The Rhetorical Framing of "Why" vs. "How")​

Explanation 1​

Explanation 2​

Explanation 3​

Explanation 4​

Explanation 5​

Task 4: AI Literacy in Practice - Reframing Anthropomorphic Language​

Task 5: Critical Observations - Structural Patterns​

Agency Slippage​

Metaphor-Driven Trust Inflation​

Obscured Mechanics​

Context Sensitivity​

Accountability Synthesis​

Conclusion: What This Analysis Reveals​

Mechanism of the Illusion:​

Material Stakes:​

AI Literacy as Counter-Practice:​

Path Forward​

Extended Processing Summary​

Task 1: Metaphor and Anthropomorphism Audit

1. Computational Matrix as Emotional State

2. Mathematical Distance as Moral Reasoning

3. Algorithmic State as Internal Feeling

4. Controller Module as Organic Brain

5. Statistical Correlation as Intentional Inference

6. Algorithmic Output as Social Collaboration

7. String Parsing as Expert Evaluation

8. Interface State as Embodied Presence

Task 2: Source-Target Mapping

Mapping 1: Biological organism with conscious emotional experience → A software architecture maintaining mathematical state vectors

Mapping 2: Ethical philosopher or conscious moral agent → Hard-coded conditional logic and state-transition rules

Mapping 3: Conscious subjective emotional states (feelings) → Initialization of a floating-point array in computer memory

Mapping 4: Biological command center (the human brain) → A basic software state-machine or progress-tracking script

Mapping 5: Psychologist or empathetic listener reading human minds → Statistical text classification API

Mapping 6: Human peer or mentor engaging in shared social work → Sequential interaction between a user and an automated text generator

Mapping 7: Expert human evaluator making conscious academic judgments → LLM text generation constrained by a prompt template

Mapping 8: Physical human being occupying physical space → A 3D graphical animation rendered on a screen

Task 3: Explanation Audit (The Rhetorical Framing of "Why" vs. "How")

Explanation 1

Explanation 2

Explanation 3

Explanation 4

Explanation 5

Task 4: AI Literacy in Practice - Reframing Anthropomorphic Language

Task 5: Critical Observations - Structural Patterns

Agency Slippage

Metaphor-Driven Trust Inflation

Obscured Mechanics

Context Sensitivity

Accountability Synthesis

Conclusion: What This Analysis Reveals

Mechanism of the Illusion:

Material Stakes:

AI Literacy as Counter-Practice:

Path Forward

Extended Processing Summary