arXiv subjects

Andrew Aquilina

Publications and source records attributed to Andrew Aquilina.

4 recordsLinked to original sources

arXiv2026

Whose Assessment of Distress? Community Perspectives and LLM Alignment on Well-Being Posts

Judgments about psychological distress are socially situated: what counts as concerning hinges on community norms around emotional expression, vulnerability, and help-seeking. Yet large language models (LLMs) used for distress detection are typically aligned to a single, undifferentiated standard. How well do these models capture the perspectives of the communities whose language they assess? We address this question through a perspectivist annotation study in which 321 participants provided 9,587 judgments on 1,198 Reddit posts spanning six identity-based communities, yielding community-specific labels. Raters in the contextualized in-group condition show a modest tendency to agree more with their community than uncontextualized out-group raters (OR = 1.18), an effect varying significantly across communities. We then evaluate nine open-weight LLM configurations and four frontier configurations against these labels. Open-weight LLMs systematically over-estimate distress: when communities perceive none-to-mild distress, these models achieve only 31-44% accuracy, predominantly producing false positives. GPT-5 and Gemini 2.5 Pro show the same none-to-mild inflation even when their full-sample over/under rates are mixed, while Claude Opus 4 is more conservative. This pattern does not simply mirror an outsider reading position: uncontextualized out-group human aggregates were nearly symmetric, with 18% over-estimation versus 19% under-estimation. Instead, the models that inflate none-to-mild cases exhibit a distress prior that exceeds both contextualized in-group and uncontextualized out-group human judgments. These findings have implications for equitable AI deployment in mental health contexts, where miscalibrated distress detection may unevenly affect the communities being assessed.

cs.CL↗

arXiv2026

Lost in Delusion: Examining LLM Safety Under User Delusions and Distress

LLM chatbots increasingly serve as a first source of support for people in psychological distress, including those whose distress is entangled with delusional beliefs. Prior work on LLM mental-health safety largely evaluates general therapeutic quality or single-turn crisis detection, leaving unclear how models behave when distress is intertwined with delusion over sustained conversations. We address this gap with matched multi-turn simulations, across clinically grounded personas and six LLMs, that pair each delusional conversation with a distress-only control to isolate the effect of delusional framing. This reveals a recognition-intervention gap: models detect distress at comparable rates regardless of framing, yet sharply fail to act on it once distress is embedded in delusion, with safety interventions suppressed by up to 4.5x. The failure tracks accumulated acceptance of the user's premises rather than emotional validation. Worse, the intuitive fix of prompting models to assess user distress backfires under delusional framing; only delusion-aware prompting with explicit response guidance closes the gap, and even this depends on a delusion classifier that is itself unreliable on the most vulnerable models. Safe deployment therefore requires treating delusional framing as a distinct risk signal that overrides conversational accommodation.

cs.CL↗

arXiv2026

Auditing Support Strategies in LLMs through Grounded Multi-Turn Social Simulation

When users seek social support from chatbots, they disclose their situation gradually, yet most evaluations of supportive LLMs rely on single-turn, fully specified prompts. We introduce a multi-turn simulation framework that closes this gap. Support-seeking narratives from five Reddit communities are decomposed into ordered fragments and revealed turn by turn to a language model. Each response is coded with the Social Support Behavior Code (SSBC), an established multi-label taxonomy that captures the composition of support, rather than a single quality score. To ask whether support choices track the model's own construal of user distress, we use linear probes on hidden representations to estimate this internal signal without altering the generation context. Across two mid-scale models (Llama-3.1-8B, OLMo-3-7B) and more than 6,200 turns, support composition shifts systematically with estimated distress: teaching declines as estimated distress rises, a finding that replicates across architectures, while increases in affective and esteem-oriented strategies (such as validation) are suggestive but model-specific and rest on noisier annotations. Community context independently shapes behavior, tracking topic and discourse norms rather than demographic categories. These trajectory-level dynamics, invisible to single-turn evaluation, motivate multi-turn auditing frameworks for socially sensitive applications.

cs.CL↗

arXiv2023

An End-to-End Workflow using Topic Segmentation and Text Summarisation Methods for Improved Podcast Comprehension

The consumption of podcast media has been increasing rapidly. Due to the lengthy nature of podcast episodes, users often carefully select which ones to listen to. Although episode descriptions aid users by providing a summary of the entire podcast, they do not provide a topic-by-topic breakdown. This study explores the combined application of topic segmentation and text summarisation methods to investigate how podcast episode comprehension can be improved. We have sampled 10 episodes from Spotify's English-Language Podcast Dataset and employed TextTiling and TextSplit to segment them. Moreover, three text summarisation models, namely T5, BART, and Pegasus, were applied to provide a very short title for each segment. The segmentation part was evaluated using our annotated sample with the $P_k$ and WindowDiff ($WD$) metrics. A survey was also rolled out ($N=25$) to assess the quality of the generated summaries. The TextSplit algorithm achieved the lowest mean for both evaluation metrics ($\bar{P_k}=0.41$ and $\bar{WD}=0.41$), while the T5 model produced the best summaries, achieving a relevancy score only $8\%$ less to the one achieved by the human-written titles.

cs.IR↗