{"id":"861b7153-509b-424f-8298-b4671bf05973","arxiv_id":"2505.10454","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"The paper introduces a three-stage emotion-sensitive grounding model (arousal, understanding, agreement) for XAI systems to adapt explanations to users' emotional states.","lead":"This paper proposes a three-stage model for emotion-sensitive explanation in explainable AI, where systems monitor a user's emotional arousal, check understanding, and seek agreement before continuing. It offers a conceptual framework for building XAI systems that adapt explanations to a user's emotional state instead of assuming rational users.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Even with perfect arousal sensing, the model's emotional module outputs only a binary anomaly flag, with no mapping from detected emotion to any specific explanation strategy; the claimed dynamic adaptation is therefore unsupported.","rationale":"The reader identified unreliable real-time emotion detection as the weakest assumption. My reading goes one step further: even if the sensing pipeline were completely reliable, the implemented emotional module reduces all emotional states to a binary z-score anomaly, and Phase Control's only response is to initiate the same clarification dialog. The model therefore cannot support the headline claim of dynamic adaptation to users' emotional states unless a mapping from specific emotional states to specific explanation strategies is specified. This is a conceptual gap, not merely an engineering concern, and it reinforces the reader's conditional verdict without requiring a change. The Discussion explicitly states that future work will evaluate effectiveness, which is an honest admission that the claimed benefit has not been demonstrated. A secondary concrete issue is that the cited EmoNet reference [4] is a speech emotion recognition framework, whereas the text claims facial-expression detection via EmoNet; this mismatch should be corrected. These observations do not make the framework internally contradictory, but they do mean the paper remains a proposal whose central mechanism is under-specified, so the existing CONDITIONAL verdict is appropriate.","tokens_in":4576,"tokens_out":4614,"duration_ms":51330,"concrete_test":"On a labeled corpus of user reactions during explanation dialogs, or via a think-aloud study, compute the Section 3 z-score trigger (window = 500 ms, threshold = 2.5) for each reaction and compare the trigger's distribution across labels such as confusion, frustration, surprise/insight, and agreement. If the binary trigger does not separate comprehension difficulties from non-problematic affective reactions, or if the model provides no mechanism to use finer-grained labels, the central claim of emotion-adaptive explanation strategy selection fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3's emotional module is a rolling z-score anomaly detector (threshold 2.5, 500 ms window) that emits a single binary deviation flag. Phase 1 either branches to the same clarification dialog or continues to the next feature. The model therefore does not encode what the user's emotional state is (valence, arousal direction, emotion category) and does not map any detected state to a choice among the \"range of different explanation strategies\" mentioned in the introduction and Section 3. The Discussion invokes an inverted-U arousal-performance curve, but a two-sided anomaly threshold treats high and low arousal as the same trigger, so the system would interrupt equally at moments of insight, frustration, and confusion. This gap is not merely an implementation detail: it is the missing link between the claimed dynamic adaptation to emotional states and the implemented Phase Control. The Discussion's admission that \"future work will evaluate the model's effectiveness\" further marks the claim as unsupported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a three-stage emotion-sensitive explanation model for XAI: (1) detection of emotional or epistemic arousal, (2) verification of understanding, and (3) assessment of agreement. The model is intended to ground explanations in a multimodal interaction with a virtual robot, using facial expression and heart-rate signals to detect deviations via a rolling z-score anomaly detector, and to initiate clarification dialogues when deviations occur. The authors argue that this provides a conceptual basis for XAI systems that dynamically adapt explanation strategies to users' emotional states, ultimately supporting better and more user-centered decision-making.","tokens_in":4862,"tokens_out":3860,"duration_ms":38302,"significance":"If fully realized, the model would address a recognized gap in XAI research: most explanation frameworks assume a rational user and do not account for emotional or epistemic arousal during explanation processing. The grounding-based framing grounded in prior work on human communication is a useful conceptual starting point, and the explicit integration of both emotional and epistemic states is commendable. However, the current manuscript is a conceptual sketch rather than a validated model. The described implementation does not actually deliver the claimed dynamic adaptation, and key sensing details contain errors. The paper's value at this stage is as a proposal; its claims need to be either substantiated or substantially scaled back.","major_comments":[{"comment":"The emotional module outputs only a binary deviation flag: a rolling z-score anomaly detector (threshold = 2.5, 500 ms window) either triggers the same clarification dialog branch or allows the system to continue to the next feature. No valence, arousal direction, or emotion category is encoded, and no mapping from any specific detected state to a particular explanation strategy is provided. The abstract's claim that the model \"dynamically adapt[s] explanation strategies to users' emotional states\" is therefore unsupported by the described implementation. The authors should either specify a mechanism by which the type or direction of arousal selects among the \"range of different explanation strategies\" mentioned in Section 2.1, or revise the claims to describe an arousal-triggered clarification dialog without the stronger adaptation claim.","section":"Section 3, Phase 1 and Figure 2"},{"comment":"The text states that real-time emotion recognition is performed via facial expressions detected by EmoNet [4]. However, reference [4] describes EmoNet as a speech emotion recognition framework, not a facial expression detector. This is a load-bearing mis-specification because Phase 1's sensing modality is a core part of the proposed model. The authors must correct the citation or the described method, or clarify which actual affect-recognition tool is intended for facial expression analysis.","section":"Section 3, Emotional Model"},{"comment":"The discussion invokes the inverted U-shaped relationship between arousal and cognitive performance, stating that the system \"targets the optimal arousal window for effective understanding.\" But the anomaly detector uses a two-sided threshold (|z| > 2.5), which treats high and low arousal as the same trigger: both initiate the same clarification dialog. Thus the model cannot distinguish over-arousal from under-arousal and cannot selectively steer users toward an optimal window without further differentiation. The inverted-U framing is therefore inconsistent with the implemented detection logic. Please either add a mechanism that distinguishes arousal direction and types (e.g., separate thresholds or classification of the anomaly sign) or temper the discussion to reflect what the model actually does, which is to react to any deviation.","section":"Section 4, Discussion"}],"minor_comments":[{"comment":"References [9] and [10] appear to refer to the same manuscript with slightly different author spellings and publication status; please reconcile and provide a single complete citation. Also, the grounding theory is attributed to \"Clark\" in the text without a full reference; please add a proper citation.","section":"References"},{"comment":"There is a typo in the first paragraph: \"purly logical considerations\" should be \"purely logical considerations.\" In Section 2, \"so calledcognitive forcing functions\" is missing a hyphen and space; it should be \"so-called cognitive forcing functions.\" ","section":"Introduction"},{"comment":"In the text, \"see fig 2\" should be capitalized and formatted as \"see Fig. 2.\" Also, the figures are referenced but not visible in the provided manuscript; please ensure they are legible and correctly placed.","section":"Figure 2"},{"comment":"The specific values of the z-score threshold (2.5) and window (500 ms) are presented as design choices without justification or citation. If these are provisional placeholders, please state so explicitly; otherwise, provide a rationale for their selection.","section":"Section 3, Emotional Model"},{"comment":"The phrase \"users emotional states\" is missing an apostrophe; it should be \"users' emotional states.\" ","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The paper is currently more of a position paper than a fully specified and validated model. The authors should be encouraged to either (a) implement and validate the proposed strategy-selection mechanism, or (b) explicitly reframe the contribution as a conceptual framework with clear limitations. The duplicate references [9]/[10] and the missing Clark citation give an impression of insufficient proofreading. If the journal's bar for acceptance includes empirical validation of proposed interactive systems, the lack of evaluation here is a substantial concern."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First: this is a framework paper, not a validated system. Second: the central weakness is not the lack of empirical evaluation; it's that the implemented emotional module does not actually adapt anything. Phase 1 produces a binary anomaly flag via a rolling z-score, and whether it fires or not, the next step is the same clarification dialog. There is no mapping from any specific emotional state (valence, arousal direction, category) to any of the several explanation strategies the paper mentions. The inverted-U arousal-performance discussion even undercuts the design, since a two-sided threshold treats insight and frustration identically.\n\nWhat's good: placing emotional and epistemic arousal at the first level of a grounding hierarchy is a sensible move, and I don't know of prior work that does this from the understanding perspective. The paper is clearly written, honestly cites its own prior results, and situates itself well in the XAI-for-decision-making literature. The three-stage model (arousal, understanding, agreement) is easy to grasp and a reasonable template for future systems. The authors are upfront that it hasn't been evaluated.\n\nSoft spots, in proportion: the binary flag issue is the main one, and it is conceptual, not an implementation detail. If you fix it by outputting a graded arousal estimate and tying it to strategy selection, the framework becomes much more credible. The detection parameters (threshold 2.5, 500ms) are stated without justification, which is a minor issue but load-bearing. No empirical evaluation is fine for a proposal, but then the abstract's claim about supporting decision-making should be softened. Circularity is not a concern: the self-citations motivate the model, not force it.\n\nWho this is for: people working on affective XAI or adaptive explanation interfaces. It's a useful concise reference, but I'd want to see the missing mapping addressed before building on it. Does it deserve peer review? Yes—it's a serious, readable proposal with a testable structure. I would send it out, with the expectation of major revision or a narrowed scope.","headline":"A clear conceptual proposal whose main gap is that the emotional module emits only a binary alarm and never maps emotion to a strategy; still worth engaging for affective XAI researchers.","tokens_in":5184,"tokens_out":2446,"would_cite":true,"duration_ms":22833,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Emotion-aware AI explanations trigger on user arousal","keywords":["explainable AI","emotion-sensitive explanations","grounding","arousal detection","human-agent interaction","clarification dialog","explanation understanding","decision support"],"falsifier":"Run a controlled study in which users receive explanations while their facial expressions and heart-rate variability are recorded, and compare system-triggered clarification events against independently rated states of confusion or understanding; if the z-score detector's triggers do not coincide with genuine comprehension difficulties (measured by follow-up comprehension questions), or if triggering clarification dialogs yields no comprehension gain over a no-trigger baseline, the emotion-sensitive grounding loop has no empirical benefit.","tokens_in":4394,"feed_emoji":"🤖","tokens_out":5920,"duration_ms":48240,"temperature":0.7,"pith_summary":"This paper argues that explainable AI systems cannot treat users as purely rational agents, because emotions influence how explanations are understood. It proposes a three-stage emotion-sensitive explanation model: detect emotional or epistemic arousal, verify understanding, and then assess agreement. The model monitors the user's facial expressions and heart-rate variability in real time; a deviation from baseline arousal triggers a clarification dialog before the explanation proceeds. The paper's contribution is a conceptual and architectural grounding for XAI systems that adapt explanatory strategy to the user's emotional state, with empirical validation left to future work.","feed_headline":"Emotion-aware AI explanations trigger on user arousal","feed_subtitle":"A three-stage model uses facial expressions and heart rate to verify understanding before seeking agreement.","key_machinery":"The central mechanism is the emotion-sensitive grounding hierarchy, a transfer of Clark-style grounding to affective human-agent interaction. It is realized by an emotional model that detects arousal via EmoNet facial-expression recognition and smartwatch heart-rate variability, using a rolling z-score anomaly detector (threshold 2.5 over a 500 ms window) to spot deviations; a cognitive model that performs risk assessment and hosts an LLM-based understanding dialog with the robot Floka; and a phase-control component that orchestrates the transitions from arousal to understanding to agreement, allowing extra phases to be inserted.","core_discovery":"The central claim is that grounding theory, originally developed for spoken-language systems, can be transferred to human-explainer interaction and extended with emotion-sensitive steps. The authors adapt the eight-level grounding hierarchy of a spoken-language telephone system to the human explainee, adding three specific stages: emotional or epistemic arousal, understanding, and agreement. At the first stage, the system watches for micro-reactions such as irritation or surprise; when a deviation is detected, it initiates a clarification sub-dialog to verify understanding, and only after understanding is established does it assess whether the user agrees with the explanation. This makes the interaction mixed-initiative: the explaining system, not only the user, can start a clarification when it senses potential misunderstanding.","pith_inferences":["A natural testable extension is to calibrate the arousal threshold empirically: compare the rolling z-score trigger against ground-truth labels of user confusion, which the paper does not provide.","The model could be broadened beyond general arousal to distinguish epistemic emotions such as confusion, curiosity, and surprise, which likely require different explanatory responses than an undifferentiated arousal spike.","The agreement stage implies a values-sensitive design: if a user objects to a feature like gender, the system can offer a counterfactual without it, effectively making the AI's value assumptions negotiable; the paper leaves this ethical dimension implicit.","The architecture is modular enough that the phase-control component could insert additional phases (for example, a 're-explain after disagreement' loop) without changing the grounding hierarchy, a flexibility the paper notes but does not explore."],"forward_implications":["XAI systems could shift from static, one-size-fits-all explanations to adaptive strategies that respond to the user's real-time emotional state.","Detecting subtle arousal signals lets the system initiate clarification dialogs on its own, enabling true mixed-initiative explanation rather than waiting for the user to ask.","The three-stage structure gives dialog designers a principled ordering: secure understanding before seeking agreement, and treat emotional reactions as a trigger rather than noise.","By tuning explanation delivery toward an optimal arousal window, the model provides a concrete route to mitigate the documented negative association between emotional intensity and explanation understanding.","The agreement stage explicitly opens the door to counterfactual explanations—presenting decisions without a contested feature—so the system can address biased or outdated training data."],"supporting_citations":[{"why":"It supplies the grounding hierarchy and the acceptance/understanding levels that the model adapts to the human explainee.","marker":"[2]"},{"why":"It provides the facial-expression recognition tool used for real-time emotion detection in the emotional model.","marker":"[4]"},{"why":"It shows explanations themselves can generate affect depending on task difficulty, motivating the reciprocal monitoring loop.","marker":"[5]"},{"why":"It reports the authors' earlier studies on prior and task-generated emotions that inform the three-stage design.","marker":"[9]"},{"why":"It provides the empirical evidence that emotional reactions are negatively associated with explanation understanding, the paper's motivating problem.","marker":"[10]"},{"why":"It demonstrates that arousal modulates reactions to explanations, supporting the choice of arousal as a grounding trigger.","marker":"[13]"}],"fun_headline_variants":["Emotion-aware XAI adapts to user micro-reactions","Three-stage model: arousal, understanding, agreement","Explanations tuned to emotional states via grounding","AI explanations sense irritation, spark clarifications"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that emotional or epistemic arousal can be reliably detected in real time from facial expressions and heart-rate variability by a rolling z-score anomaly detector with a 2.5 threshold over a 500 ms window, and that such detected deviations are valid triggers for clarification dialogs.","fun_headline_variants_meta":{"raw":{"variants":["Emotion-aware XAI adapts to user micro-reactions","Three-stage model: arousal, understanding, agreement","Explanations tuned to emotional states via grounding","AI explanations sense irritation, spark clarifications"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00016,"raw_usage":{"total_tokens":1140,"prompt_tokens":764,"completion_tokens":376,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":380,"completion_tokens_details":{"reasoning_tokens":315}},"tokens_in":380,"tokens_out":376,"duration_ms":4475,"temperature":1.0,"reasoning_tokens":315,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T21:07:52.222420+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a controlled study in which users receive explanations while their facial expressions and heart-rate variability are recorded, and compare system-triggered clarification events against independently rated states of confusion or understanding; if the z-score detector's triggers do not coincide with genuine comprehension difficulties (measured by follow-up comprehension questions), or if triggering clarification dialogs yields no comprehension gain over a no-trigger baseline, the emotion-sensitive grounding loop has no empirical benefit.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It supplies the grounding hierarchy and the acceptance/understanding levels that the model adapts to the human explainee."},{"cited_title":"Gerczuk, S","cited_arxiv_id":null,"evidence_quote":"It provides the facial-expression recognition tool used for real-time emotion detection in the emotional model."},{"cited_title":"Guerdan, A","cited_arxiv_id":null,"evidence_quote":"It shows explanations themselves can generate affect depending on task difficulty, motivating the reciprocal monitoring loop."},{"cited_title":"Influence of prior and task generated emotions on XAI explanation retention and understanding","cited_arxiv_id":"2505.10427","evidence_quote":"It reports the authors' earlier studies on prior and task-generated emotions that inform the three-stage design."},{"cited_title":"Richter, C","cited_arxiv_id":null,"evidence_quote":"It provides the empirical evidence that emotional reactions are negatively associated with explanation understanding, the paper's motivating problem."},{"cited_title":"Thommes, O","cited_arxiv_id":null,"evidence_quote":"It demonstrates that arousal modulates reactions to explanations, supporting the choice of arousal as a grounding trigger."}],"review_version":1}