PaperlayerView exact revision on arXiv

Source-linked AI summary

When a Story Feels Like Mine: How Personalized Narratives and Humor Shape Older Adults' Empathy toward LLM-Generated Peer Health Stories

Kexin Quan, Precious Olalere, Smit Desai, Jessie Chin

arXiv:2609.16374v2cs.HC

TL;DR

Evidence is limited on whether situational personalization and humor jointly shape older adults’ responses to LLM-generated peer health stories. The paper develops a self-efficacy-grounded pipeline for generating personalized narratives and tests four conditions with older adults. Personalization improved perceived relevance and relatability, while humor showed no reliable benefit for perceived similarity.

  • Problem

    Whether situational similarity in LLM-generated peer health narratives strengthens older adults’ empathy remains unclear, and matching diverse barriers and coping styles is a challenge for personalizing these stories.

    Source: §1 · p. 2

  • Method

    A three-stage, self-efficacy-grounded LLM pipeline turns older adults’ reported health-habit challenges into personalized first-person peer stories, tested in a four-condition within-subject study.

    Source: §1 · p. 2, §3.2 · p. 6

  • Results

    Personalized peer health narratives improved perceived relevance and relatability, while humor produced no reliable benefit for perceived similarity.

    Source: §7 · p. 13

  • Takeaways & Limitations

    The findings support prioritizing personally meaningful situations and actionable turning points while evaluating readers’ responses to affective style.

    Source: §7 · p. 13

    The small, demographically skewed sample, ceiling effects, narrow humor-preference range, and stimuli that varied in multiple narrative features constrain the findings.

    Source: §6.6 · p. 13

Abstract

from arXiv · show

Peer stories have been shown to boost self-efficacy in older adults' health behavior change. Despite their effectiveness, peer stories are difficult to deploy in health promotion at scale given the difficulty of matching the diverse health concerns and coping styles of heterogeneous older populations. Large language models (LLMs) have been shown to generate authentic narratives, yet how personalization and narrative affective style, such as humor, jointly shape older adults' responses remains unknown. We developed a theory-driven system that generates first-person peer health narratives varying in personalization and humor through a three-stage LLM pipeline grounded in self-efficacy mechanisms. Thirty-one older adults were invited to participate in a within-subjects lab study. Results showed that personalization increased perceived relatability and relevance of peer stories, especially for older adults with lower humor preference. These findings position individual differences in affective styles as a second dimension in designing personalization for LLM-assisted health communication.

1 INTRODUCTION

The paper asks how situational personalization and humor jointly shape older adults’ responses to peer health narratives. It introduces a self-efficacy-grounded LLM system and reports that personalization increases perceived relevance and relatability, with responses varying by affective style.

1 INTRODUCTION

  • Whether situational similarity in LLM-generated peer stories strengthens older adults’ empathy remains unclear, despite the importance of matching stories to readers’ barriers and fitting coping strategies.People with the same health concern can face different behavior-change barriers, so profile matching alone may not provide a relatable experience.Source: §1 · p. 2
  • Narrative content sets the experience portrayed, while tone shapes its emotional presentation; humor may foster connection or reduce perceived seriousness and invite message discounting.The paper examines personalization and humor together to distinguish situational fit from narrative tone.Source: §1 · p. 2
  • The study generated first-person peer stories from older adults’ reported challenges and evaluated four personalization-by-humor conditions with 31 community-dwelling older adults.Participants rated empathy, understanding, relatability, relevance, and authenticity; the study also collected preferences and verbal reflections.Source: §1 · p. 2
  • Personalization increased perceived relevance and relatability, with effects varying by affective style; humor made stories more engaging but did not by itself strengthen empathy or relatability.The contribution statement identifies humor preference as one individual difference in affective style.Source: §1 · p. 2
  • The authors frame scalable health communication around personally relevant peer narratives paired with actionable strategies intended to support older adults’ self-efficacy.The system’s stories reflect participants’ barriers and goals and embed coping strategies matched to their needs.Source: §1 · pp. 2–3

2 RELATED WORK

Prior work motivates narrative-based health communication for older adults, personalization around lived circumstances, and humor tailored to audience and context. This study builds on those strands to examine situational fit and humor together in peer health stories.

2.1 Storytelling as a Health Communication Medium for Older Adults

  • Stories place health information in a character’s experiences; prior work connects narratives with processing, social connection, reduced resistance, and state empathy, while older adults’ storytelling practices include life review.Research on older adults’ storytelling also describes psychosocial benefits from reminiscence interventions and storytelling.Source: §2.1 · p. 3

2.2 Personalized Communication and Empathic Engagement in Later Life

  • Prior work links similarity and alignment with identification, persuasion, and self-efficacy, while evidence on LLM-generated peer health narratives for older adults remains limited.Older adults are reported to value personalization and content congruent with lived experience; the paper therefore examines personalization alongside humor.Source: §2.1 · p. 3, §2.2 · p. 3

2.3 Humor Reception in Health Messages Across Individuals and Age

  • Humor’s effects depend on style, recipient inclination, and age-related comprehension; affiliative humor is generally more adaptive, and inauthentic humor can be counterproductive.Prior studies leave open whether humor strengthens personalized AI-generated peer health narratives for older adults and which readers are most receptive.Source: §2.3 · p. 4

3 SYSTEM DESIGN

The system crosses personalization and humor to deliver four first-person peer health stories per participant. Personalized narratives map reported health-habit challenges to a support need, strategy, and story, while the shared template and humor prompts constrain their form and tone.

3 SYSTEM DESIGN

  • The system delivers four LLM-authored stories per participant, crossing two fixed or personalized narratives with humor on or off; the study procedure has participants read and compare all four.Personalized stories are generated from each participant’s pre-survey, while fixed stories are reused across participants.Source: §3 · p. 4, Figure 2 · p. 4

3.1 Shared Narrative Template

  • All conditions use a 300–375-word, three-act first-person template: an everyday struggle, a strategy-centered turning point, and partial but meaningful progress.Prompts also request a warm, emotionally believable tone and exclude clinical, prescriptive, inspirational, or overly polished language.Source: §3.1 · pp. 4–5

3.2 Self-Efficacy-Guided Personalization of Peer Narratives

  • The three-stage personalization pipeline infers an efficacy support need from participant responses, selects a fitting behavior-change strategy, then makes that strategy the narrative turning point.The library organizes 36 literature-derived strategies into six categories; personalization prioritizes similar everyday difficulties, barriers, and values over demographic matching.Source: §3.2 · pp. 5–6

3.3 Humor Manipulation

  • Humor prompts specify affiliative, coping-oriented humor from everyday routines and gentle self-directed observations, while excluding sarcasm, ridicule, teasing, and culture-specific references.Non-humorous prompts instead call for a quiet opening in which the narrator notices the problem.Source: §3.3 · p. 6

4 STUDY DESIGN

The study used a 2 × 2 within-subject design in which older adults rated four peer narratives. Its procedure combined repeated story ratings and forced-choice comparisons, with quantitative models and inductive thematic analysis of verbal responses.

4 STUDY DESIGN

  • A 2 × 2 within-subject study examined how personalization, humor, and their interaction affected older adults’ evaluations of peer health narratives.Each participant evaluated all four conditions.Source: §4 · p. 6

4.1 Participants

  • Thirty-one community-dwelling older adults aged 61–82 completed the study after recruitment through community and university channels.Participants received $35 for the full laboratory session; the university IRB approved the voluntary study.Source: §4.1 · p. 6

4.2 Study Procedure

  • The scheduled 30-minute procedure comprised a pre-survey, four story readings with ratings, a comparison survey, and optional verbal questions.Participants rated five outcomes after each story and then chose stories for five forced-choice outcomes; stories were presented in one of eight predefined orders.Source: §4.2 · pp. 6–7

4.3 Statistical and Thematic Analysis Method

  • Linear mixed-effects models tested personalization, humor, and interaction effects on five ratings; inductive thematic analysis examined verbal responses, with final coding decisions made by the researcher.The quantitative analyses included sensitivity checks and exploratory moderation models; a researcher reviewed and refined the collaboratively proposed qualitative codes and themes.Source: §4.3 · pp. 7–8

5 RESULTS

Personalization increased perceived relevance and relatability, while humor alone did not shift narrative ratings. Personalization effects were larger among older adults with lower humor preference.

5.1 Data Completeness and Descriptive Results

  • All 31 participants rated four stories, and all 124 narratives met the 300–375-word requirement; condition means across ratings ranged from 5.77 to 6.71.Ratings covered five per-story measures on seven-point Likert scales.Source: §5.1 · p. 8, Table 2 · p. 8

5.2 Norming Study: Manipulation Check on Humor Expression

  • β=1.95 on the 0–10 scale was the estimated humor-rating difference between humorous and non-humorous stories (95% CI [1.54, 2.36], p < .001).The manipulation norming study included 64 narratives rated by 12 raters; 29 of 31 personalized pairs had higher mean ratings for the humorous version.Source: §5.2 · pp. 8–9

5.3 RQ1: Personalized Stories Received Higher Relevance and Relatability Ratings

  • 0.58 points higher for relevance and 0.40 points higher for relatability were the ratings for personalized versus non-personalized stories.The differences had 95% CI [0.26, 0.90], pHolm = .002 for relevance and CI [0.09, 0.72], pHolm = .047 for relatability; empathy, understanding, and authenticity were equivalent across narratives.Source: §5.3 · p. 9, Figure 5 · p. 9
  • Readers linked relevance and relatability to recognizing their own struggles; 20 of 29 described recognizing their behavioral patterns, and 18 discussed strategy usefulness.Only eight participants explicitly identified personalization.Source: §5.3 · p. 9

5.4 RQ2: Humor Showed No Effects on Perceived Relevance, Relatability, Understanding, Empathy, and Authenticity of Peer Stories

  • Humor had no significant main effects on empathy, understanding, relatability, relevance, or authenticity, and it did not interact with personalization on these perceptions.Participants described humor as engaging, but familiarity of the situation remained more important to empathy; some also reported fatigue with the story style.Source: Figure 5 · p. 9, §5.4 · p. 10

5.5 RQ3: Older Adults with Lower Humor Preference Demonstrated Larger Effects of Personalization on Relevance and Relatability of Peer Stories

  • β=−0.248 for relatability and β=−0.231 for relevance indicated larger personalization effects among older adults with lower humor preference; both interactions survived FDR correction (pFDR = .042).No other moderation effects survived FDR correction, including those involving humor engagement, humor attention, or age.Source: §5.5 · p. 10
  • Story judgments also depended on fit: one participant preferred a larger action, putting air in bicycle tires, over merely checking tire pressure.Another participant described their personalized non-humorous story as genuine and realistic, while calling other stories forced.Source: §5.5 · p. 10

6 DISCUSSION

Personalization most clearly helped older adults recognize their own struggles and plausible coping strategies in peer stories, while humor’s benefits depended on its fit with the story and reader. The discussion proposes iterative, context-sensitive design while noting ethical concerns and limits on what the study establishes.

6.1 RQ1: Recognizing One’s Own Situation in a Peer Narrative

  • Personalization most clearly improved perceived content fit through relevance and relatability, while evidence for greater empathy toward the narrator was mixed.Participants connected stories to familiar barriers and imaginable actions; the rating effect on empathy did not survive correction.Source: §6.1 · pp. 10–11

6.2 RQ2: Humor’s Appeal and Narrative Fit

  • Humor’s appeal as a reading experience did not translate into detectable benefits for perceived relevance, relatability, empathy, understanding, or authenticity.Participants described both enjoying humorous stories and tiring of their recurring style; amusement and empathic connection could depend on different narrative features.Source: §6.2 · p. 11

6.3 RQ3: Individual Differences in Responses to Personalization

  • Lower humor preference was associated with larger personalization effects on relevance and relatability, while the personalization-by-humor interaction remained inconclusive.The discussion describes these moderation findings as exploratory and notes generally positive humor attitudes among participants.Source: §6.3 · pp. 11–12, Figure 6 · p. 11

6.4.1 Capture Behavior-Change Context at Onboarding and Refine Preferences Through Use.

  • Twenty-six of 31 participants chose personalized stories as most relevant; the authors propose onboarding about behavior-change context and separate feedback on content fit and tone.Readers should be able to revise their circumstances and preferences as they use the system.Source: §6.4.1 · p. 12

6.4.2 Make the Next Step Meaningful.

  • A strategy can be easy to perform yet offer little meaningful progress, so narrative turning points should reflect readers’ capabilities, prior attempts, and desired outcomes.The authors propose asking what readers already do, have tried, and where progress stalls, then using their feedback to guide strategy selection.Source: §6.4.2 · pp. 12–13

6.5 Ethical Considerations

  • Relatable peer progress may lend authority to an unsuitable strategy; the study’s prompts do not establish clinical safety or effectiveness.The authors call for confidentiality protections and preserving readers’ freedom to accept, adapt, or reject depicted actions.Source: §6.5 · p. 13

6.6 Limitations and Future Work

  • The findings are bounded by a small, demographically skewed sample, limited humor-preference variation, non-isolated narrative features, and single-session, single-item measures.Future work should use more diverse samples, more tightly matched stimuli, direct measures of humor perception, and longitudinal or behavioral outcomes.Source: §6.6 · p. 13

7 CONCLUSION

In a study with 31 older adults, personalized peer health stories improved perceived relevance and relatability, while humor showed no reliable benefit for perceived similarity. The conclusion supports prioritizing meaningful situations and actionable turning points while evaluating readers’ responses to affective style.

7 CONCLUSION

  • In a within-subject study with 31 older adults, LLM-personalized peer health narratives improved perceived relevance and relatability, while humor produced no reliable benefit for perceived similarity.A three-stage pipeline incorporated health-habit difficulties and practical strategies as narrative content.Source: §7 · p. 13

A.1 Per-Story Ratings

The appendix describes the per-story ratings and pre-survey measures used to collect participants’ responses and health-habit context.

A.1 Per-Story Ratings

  • After each story, participants rated five statements on seven-point agreement scales; all participants completed every item for all four stories.The measure labels correspond to those used in the results.Source: §A.1 · p. 17, Table 3 · p. 17

A.2 Pre-Survey Humor Measures

  • The pre-survey measured humor preference, humor engagement, and humor attention using three seven-point agreement items.The items asked about enjoying light humor, engaging with difficult topics, and staying interested in humorous stories.Source: §A.2 · p. 17

A.3 Pre-Survey Health-Habit Items

  • Five personalization-input items asked about a health habit and current stage, including starting, maintaining, or restarting it, plus difficulties and setbacks.One item appeared in a version conditional on the participant’s selected stage.Source: §A.3 · pp. 17–18
  • The health-habit items also asked why the change mattered and what the participant hoped it would help them do, continue, or regain.Source: §A.3 · p. 18

A.4 Post-Survey

After reading the four stories, participants made forced choices and answered open-ended questions about empathy, fit, authenticity, and their reading experience.

A.4 Post-Survey

  • After all four stories, participants selected one story for each of five forced-choice questions about empathy, relatability, relevance, authenticity, and interest in more stories.Source: §A.4 · p. 18
  • Table 4 reports forced-choice selections by factor, with exact binomial tests against 50% chance; humor counts are shown as ranges for two possible codings.Source: Table 4 · p. 18
  • Participants also answered three verbal open-ended questions about why their chosen story elicited empathy, what made stories relatable or relevant, and other reading experiences.Source: §A.4 · p. 18

B INDEPENDENT HUMOR CHECK ASSIGNMENT

The humor check distributed fixed and personalized story versions across raters in randomized blocks, yielding 264 ratings overall.

B INDEPENDENT HUMOR CHECK ASSIGNMENT

  • The 31 personalized story pairs each had humorous and non-humorous versions generated from the same participant profile.The check included 64 unique narratives and two fixed stories forming the non-personalized pair.Source: B INDEPENDENT HUMOR CHECK ASSIGNMENT · p. 18
  • Each of 12 raters evaluated 10 personalized pairs and two fixed stories in a 22-story block.Three personalized pairs were shared across raters; seven were drawn from the remaining 28 pairs.Source: B INDEPENDENT HUMOR CHECK ASSIGNMENT · p. 18
  • 264 ratings covered 240 personalized-story ratings and 24 ratings of two fixed stories.Each fixed story was rated by all 12 raters.Source: B INDEPENDENT HUMOR CHECK ASSIGNMENT · p. 19
  • Story order was randomized, paired versions were separated, and formatting was held constant across conditions.Raters scored each story’s perceived funniness and lightheartedness from 0 to 10.Source: B INDEPENDENT HUMOR CHECK ASSIGNMENT · p. 19

C BEHAVIOR-CHANGE MECHANISM LIBRARY

The curated mechanism library supplies concrete behavioral strategies organized by self-efficacy support needs, with six examples per category for Stage 2.

C BEHAVIOR-CHANGE MECHANISM LIBRARY

  • The library contains 36 mechanisms across six self-efficacy support-need categories, with six examples per category supplied to Stage 2.The model could select an example or generate a better-fitting alternative; each example describes a concrete action suited to a narrative turning point.Source: C BEHAVIOR-CHANGE MECHANISM LIBRARY · p. 19, Table 5 · p. 20
  • Initiation strategies include micro-steps, if–then plans, scaled-back starts, habit stacking, realistic self-assessment, and a low-threshold first win.The examples target accessible action and early confidence.Source: C BEHAVIOR-CHANGE MECHANISM LIBRARY · p. 20
  • Maintenance examples reduce friction or support consistency through environment design, weekly check-ins, streak tracking, and a two-day rule.Other listed strategies include friction removal and identity-based goals.Source: C BEHAVIOR-CHANGE MECHANISM LIBRARY · p. 20
  • Coping strategies include Plan B activities, indoor alternatives, active pacing, preemptive coping, self-compassion, and token versions of habits.These address pain, fatigue, stress, or disruptions while continuing a routine.Source: C BEHAVIOR-CHANGE MECHANISM LIBRARY · p. 20
  • Consultation examples help people seek support through prioritized questions, verification calls, plain-language scripts, specific asks, and peer groups.The category addresses difficulty asking for help, clarification, or support.Source: C BEHAVIOR-CHANGE MECHANISM LIBRARY · p. 20
  • Recovery strategies support restarting with easier re-entry, lapse reframing, graded returns, trigger analysis, or a scheduled restart date.These address difficulty resuming after stopping or leaving a routine.Source: C BEHAVIOR-CHANGE MECHANISM LIBRARY · p. 20
Loading 2609.16374v2…