PaperlayerView exact revision on arXiv

Source-linked AI summary

Me Among Us: Affective Framing in Data Donation

Zeya Chen, Zach Pino

arXiv:2608.20523v1cs.HC

TL;DR

Data-donation consent remains difficult when data use is abstract, and little is known about how affective framing through visualization shapes these decisions. The study compares three framing lenses in a real-world calendar-data donation study and finds that the individual-collective lens produced the most favorable overall response, while framing activated distinct affective mechanisms.

  • Problem

    Text-based consent can leave data-use information abstract, and visualization-mediated affective framing in data donation remains insufficiently understood.

  • Method

    A real-world study of 24 participants compared individual-donor, individual-collective, and collective-institutional framing through institutional calendar-data visualizations and affective measures.

  • Results

    The individual-collective lens had an 87.5% donation rate and appeared to best balance personal relevance with social context, while the three frames activated distinct dominant affective mechanisms.

  • Takeaways & Limitations

    Framing visualizations to balance personal relevance with social context appears most supportive of informed, positive donation decisions, while decision-relevant affect should be supported rather than simply eliminated.

  • Takeaways & Limitations

    The exploratory study used 24 participants in a controlled setting, one relatively less-sensitive data type, and a predominantly female institutional sample.

Abstract

from arXiv · show

This study investigates how different framing approaches influence the affective aspects of data donation decision-making. Although framing effects are well studied in charitable giving, how affective framing shapes data donation, especially through data visualization, remains poorly understood. Using a theoretical framework based on the functions of affect in decision-making, we examine how three distinct framing approaches, an individual-donor lens (Group A), an individual-collective lens (Group B), and a collective-institutional lens (Group C), shape participants' affective experiences and subsequent donation decisions. Through a real-world data donation study (N=24), we found that framing designs substantially influenced donation outcomes, with the individual-collective lens generating the most favorable responses. Our analysis illustrates how affect can functions as information, motivation, and as a spotlight during the decision-making process, providing insights for designing more informed data donation interfaces and communications. This research contributes to understanding the complex interplay between framing designs, affective responses, and decision outcomes in data donation contexts.

I. INTRODUCTION

The study addresses how framing and visualization shape affective data-donation decisions, extending established framing research beyond traditional donation contexts. It examines three lenses—individual-donor, individual-collective, and collective-institutional—through theories of affect in decision-making.

  • Data donation is a voluntary, donor-centered contribution of personal data for specified common-good purposes without immediate compensation.
  • Text-based consent can leave data-use information abstract, while visualization may make complex data practices more concrete and comprehensible.
  • Little is known about how visualization-mediated framing affects data-donation decisions involving privacy, trust, and self-disclosure.
  • The study compares individual-donor framing focused on personal insights, individual-collective framing emphasizing social comparison, and collective-institutional framing emphasizing institutional benefits.
  • The framework treats affect as information that assigns value, motivation that supports goal-directed behavior, and a spotlight that shifts attention and information weighting.

III. METHOD

The study used a real-world institutional calendar-data donation scenario to compare three framing designs presented through shared interactive visualizations. Behavioral, cognitive, and affective measures were combined in a mixed-methods analysis.

  • Participants made a genuine opportunity to donate anonymized institutional calendar data to their institution, rather than responding to a hypothetical scenario.
  • The three conditions emphasized personal self-discovery, individual positioning within a collective, or integration into institutional patterns.
  • All groups viewed the same four visualization archetypes, while data comparison, tooltip text, and instructions varied by framing condition.
  • The final sample included 24 participants, equally distributed across the three framing groups, with 8 participants per group.
  • The mixed-methods analysis measured affective valence and intensity with 7-point scales and examined motivation, spotlight effects, and affective information through behavioral measures and thematic analysis.

IV. RESULTS

Framing conditions differed in donation behavior, exploration, and perceived helpfulness: Group B produced the most favorable pattern, while Group C was lower and more polarized. The donation comparison was practically sizable but not conventionally significant at N=24.

  • Donation Decision Outcomes: 62.5% of participants donated; donation rates were 62.5% for Group A, 87.5% for Group B, and 37.5% for Group C.
  • Donation Decision Outcomes: The Group B–Group C donation-rate gap was 50 percentage points, with Cramér’s V=0.42, while the omnibus and pairwise tests reported p=.12.
  • Exploration Duration: M=14.9 minutes (SD=5.30) was the average exploration time, with a significant framing difference, F(2, 21)=7.75, p=.003, η²=0.43.
  • Exploration Duration: Group B explored longest at M=19.1 minutes (SD=3.71), whereas Group C showed the greatest variance and polarized shorter-versus-extended exploration.
  • Perceived Helpfulness: Perceived helpfulness was highest in Group B (M=6.00, SD=0.93), and all eight Group B participants reported positive helpfulness ratings.

C. Affect-as-Spotlight: Attention and Information Weighting

Data exploration changed what participants understood and attended to, with framing directing attention toward different donation topics. Affective responses also differed by donation outcome and framing lens.

  • Affect-as-Spotlight: Attention and Information Weighting: Understanding how much and what types of data would be collected increased by M=2.21, while understanding who could access institute calendar data slightly decreased by M=-0.21.
  • Affect-as-Spotlight: Attention and Information Weighting: Group A showed its strongest understanding increase for data amount and types collected, with M=2.25 (SD=2.60).
  • Affect-as-Spotlight: Attention and Information Weighting: Group B showed the strongest overall changes and a M=2.75 increase in understanding what the institute could see from calendar data.
  • Affect-as-Information: Donors typically expressed curiosity, excitement, comfort with sensitivity, and appreciation of benefits, whereas non-donors more often expressed confusion, skepticism, and privacy concerns.
  • Affect-as-Information: Group A’s positive feelings were centered on self-discovery and were less directly related to donation considerations, indicating potential incidental affect bias.
  • Affect-as-Information: Group B emphasized social comparison and relationships, while Group C expressed the most mixed feelings and difficulty identifying individual positions within broader patterns.

V. DISCUSSION

Framing conditions produced different donation outcomes and affective mechanisms, with the individual-collective lens yielding the highest donation rate. The discussion positions affective framing as a design lever whose effects vary across the decision journey and whose interpretation is bounded by the study’s exploratory context.

  • Affective Framing Effect in Data Donation: 87.5% donation rate was observed for Group B, compared with 62.5% for Group A and 37.5% for Group C.The individual-collective lens produced the most favorable donation outcome among the three conditions.
  • Affective Framing Effect in Data Donation: Group A’s personal-discovery affect was often disconnected from donation, while Group B aligned engagement, understanding, and emotion with the sharing decision.Group C instead elicited confusion or disengagement and showed the smallest understanding gains.
  • Temporal Dynamics of Affect in Decision Processes: Across conditions, affective framing motivated engagement, directed attention, and generated emotions relevant to donation decisions.The findings describe affect as operating through motivation, spotlight, and information across different stages and frames.
  • Affective Framing as a Design Strategy: The individual-collective effect depended on combining collective framing with a personal anchor rather than using collective appeals alone.Group B balanced personal relevance with collective benefit, whereas Group C’s collective-institutional framing lacked that balance.
  • Affective Framing as a Design Strategy: Affective computing can treat conceptual framing as a design strategy alongside interface-level emotion detection, response, or simulation.This framing-based intervention does not require emotion detection technologies and can shape the affective journey through decision points.
  • Limitations and Future Work: The study’s conclusions are exploratory because it used 24 participants, one data type, a controlled setting, and a sample drawn from the researchers’ institution.The authors also note that institutional culture may have influenced willingness to donate.

VII. CONCLUSION

The study associates framing approaches with distinct affective dimensions of data donation, with the individual-collective lens appearing to best balance personal relevance and social context. It extends affective-computing design toward conceptual framing while emphasizing privacy, accessibility, voluntary participation, and ethical safeguards.

  • Conclusion: 87.5% donation rate was associated with the individual-collective lens, which appeared to balance personal relevance with social context.The three framing conditions activated distinct dominant affective mechanisms.
  • Conclusion: Framing visualizations to balance personal relevance with social context appeared most supportive of informed, positive donation decisions.The authors argue that decision-relevant affective engagement should be part of informed consent rather than treated only as a bias.
  • Ethical Impact Statement: The protocol addressed data-donation risks through institutional review, informed consent, privacy protections, and voluntary participation.Participants could decline donation or leave the experiment, and donated data was anonymized and securely stored.
  • Ethical Impact Statement: Accessibility accommodations included reduced motion, high contrast, an alternative color palette, redundant shape-and-color encoding, keyboard navigation, and text alternatives.The authors state that these measures supported ethical inclusivity and reduced the possibility that accessibility barriers would confound affective experiences.

B. Potential Risks and Limitations

The study identifies ethical and methodological risks involving affective manipulation, privacy during data collection, limited generalizability, and supplementary-material scope.

  • Affective Manipulation Concerns: Framing effects may be misused to maximize donation rates, so the authors prioritize informed decision-making over persuasion.They consider framing acceptable only when it improves understanding, and regard informed refusal as a valid outcome.
  • Research Experiment Logistical Concerns: The open floor-plan study setting may have made some participants feel exposed despite privacy accommodations.Mobile partitions and non-peak scheduling were used to minimize disruption and protect privacy.
  • Generalizability Limitations: The single-institution graduate-student sample limits generalizability to other populations and contexts.The authors call for replication with larger and more diverse samples across demographic and sociocultural dimensions.
  • Supplementary Materials: The supplementary materials provide additional reproducibility detail but are not required to follow the paper’s argument.Sections, tables, and figures are prefixed for cross-reference with the main text.
  • Measurement Documentation: The supplementary instruments document the measures used for demographics, privacy-risk awareness, understanding, helpfulness, and post-decision feelings.These materials connect survey items to reported analyses and tables.

A. Demographic and Background Survey

The demographic and background survey collected participants’ identity, academic, geographic, data-donation, calendar-use, and privacy-understanding information.

  • Demographics and Background: Participants reported gender identity, current program and semester, prior professional or academic background, and previous degree location.Gender responses included self-description and prefer-not-to-say options.
  • Data-Donation Experience: Prior data-donation experience was measured categorically, from no experience or knowledge to repeated participation.The survey also allowed participants to indicate uncertainty and provide an optional description.
  • Calendar Use: Calendar background questions assessed duration and frequency of Google Calendar use, knowledge of its functions, and access platforms.Access options included desktop apps, mobile apps, and integrations with other platforms.
  • Privacy and Understanding: A 7-point Likert battery assessed knowledge of calendar access, contents, collection, research purpose, processing, institutional visibility, and benefits.Additional items addressed personal benefits and potential information-exposure or privacy risks.
  • Donation Implications: The survey separately measured understanding of collective benefits, personal benefits, and potential privacy risks associated with donation.These items distinguish community-oriented and self-oriented expectations from perceived risk.

C. After-Study Survey

The after-study survey assessed donation feelings, visualization impact, perceived knowledge, decision confidence, and participants’ takeaways after data exploration.

  • Decision Feelings: Participants reported how they felt about their donation decision using categories ranging from generally good to very uncomfortable.The response set also included mixed, neutral, uncertain, and other feelings.
  • Visualization Impact: Participants rated each visualization’s impact from 0 to 5, with an additional not-applicable option.The survey distinguished event-focused dot plots, connection-focused network maps, and personalized data creation.
  • Post-Study Understanding: Post-study items assessed knowledge of calendar features, data access, contents, collection, research purpose, processing, and administrator visibility.These measures revisit informational understanding after exploration.
  • Donation Understanding: The survey measured understanding of collective benefits, personal benefits, and potential information-exposure and privacy risks.These items parallel the pre-study assessment while referring to the donation decision.
  • Decision Readiness and Takeaways: Participants indicated whether they knew enough to decide about donation and described their biggest takeaway from exploring their data.The open response asked what they learned, gained, or discovered.

S3. TASK FORMULATIONS AND THINK-ALOUD PROMPTS

The think-aloud protocol elicited participants’ observations, interpretations, feelings, and reflections on how visualizations affected donation decisions.

  • Think-Aloud Instruction: Participants were instructed to continuously describe what they noticed, understood, found confusing, and felt while exploring each visualization.The protocol emphasized that there were no right or wrong answers.
  • Nondirective Probing: Facilitators used nondirective probes when participants fell silent rather than steering them toward donation or refusal.Prompts asked what participants noticed, made of the display, or felt about it.
  • Decision and Affect Probes: Additional prompts asked whether the visualization changed participants’ donation thinking and what made them more or less comfortable donating.Affect-focused probes elicited in-the-moment expressions used in the affect-as-information analysis.

S4. AFFECTIVE CODING FOR THE AFFECT-AS-INFORMATION ANALYSIS

The affect-as-information analysis used reflexive thematic coding of participants’ affective expressions during exploration and after the study. Expressions were classified into six affective categories relevant to interpreting data-donation decisions.

  • Coding procedure: A single analyst reflexively coded think-aloud comments and open-ended post-study responses for affective expressions.The analysis focused on in-the-moment and retrospective affect; post-decision feeling ratings were handled separately.
  • Affective categories: INTEREST captured curiosity about the visualizations or interaction.
  • Affective categories: COMFORT captured confidence with the data being shown or shared.
  • Affective categories: SURPRISE captured surprise or delight at what the data revealed.
  • Affective categories: CONFUSION captured difficulty understanding the data or visualizations.
  • Affective categories: WORRY and SKEPTICISM captured concerns about privacy or exposure and doubts about data accuracy or security.

B. Integral vs. incidental marking

Affective expressions were marked integral when they concerned topics relevant to donation decisions and incidental when they did not. The groups differed in both relevance and interpretive stance, aligning with distinct affective mechanisms.

  • Integral vs. incidental marking: Integral expressions concerned donation-relevant topics, whereas incidental expressions concerned topics such as aesthetic delight or personal-memory recall.
  • Group differences: Group A’s affect was predominantly incidental, while Group B’s was predominantly integral.
  • Interpretive stance: Group A was largely self-focused, Group B comparative, and Group C collective, with some Group C confusion about whose data was shown.
  • Interpretive stance: The groups’ affective readings aligned with different dominant mechanisms: information/incidental for Group A, spotlight/integral for Group B, and motivation with high variance for Group C.
  • Visualization controls: All conditions used the same four interactive visualization archetypes, with visible differences attributable to framing design rather than underlying data.Conditions differed in data comparison, tooltip text, and on-screen instructions; the views used the same example participant’s donated calendar data.
  • Visualization controls: The exploration platform presented participants with the archetypes sequentially before the donation decision.
Loading 2608.20523v1…