Source-linked AI summary
Data is Personal: Attitudes and Perceptions of Data Visualization in Rural Pennsylvania
Evan M. Peck, Sofia E. Ayuso, Omar El-Etr
TL;DR
The paper addresses the limited evidence about how rural, economically and educationally diverse populations attend to and trust data visualizations. Through 42 interviews in rural Pennsylvania, it finds that perceptions reflect intertwined factors including clarity, visual appeal, education, political identity, and personal experience.
Problem
Visualization guidelines often derive from studies that underrepresent rural populations facing information and infrastructure inequalities.
Method
Researchers conducted 42 semi-structured interviews using ten drug-related visualizations and analyzed transcripts with grounded-theory-informed coding.
Results
Participants’ attitudes reflected a complex mix of clarity, simplicity, color, education, political identity, source perceptions, and personal experience.
Takeaways & Limitations
Designing for rural audiences requires attention to personal framings and diverse backgrounds alongside visual form and clarity.
Takeaways & Limitations
The study’s representativeness is unclear because it involved 42 participants, 10 graphs, and recruitment largely at a local farmers market.
Abstract
from arXiv · showhide
Many of the guidelines that inform how designers create data visualizations originate in studies that unintentionally exclude populations that are most likely to be among the 'data poor'. In this paper, we explore which factors may drive attention and trust in rural populations with diverse economic and educational backgrounds - a segment that is largely underrepresented in the data visualization literature. In 42 semi-structured interviews in rural Pennsylvania (USA), we find that a complex set of factors intermix to inform attitudes and perceptions about data visualization - including educational background, political affiliation, and personal experience. The data and materials for this research can be found at https://osf.io/uxwts/
1 INTRODUCTION
The paper examines whether visualization research generalizes to rural populations that face information, infrastructure, and educational inequalities. It uses interviews to identify factors shaping attention and perceptions of data visualizations.
- Rural populations may face gaps in education, income, device availability, and internet access that affect access to understandable data.
- Existing accessibility efforts often emphasize education, although rural infrastructure and funding challenges complicate large-scale literacy initiatives.
- The study reports 42 interviews with rural Pennsylvania residents to examine factors influencing perceptions of and attention toward data visualizations.
- The literature identifies experience, education, attention, emotion, familiarity, and bias as factors that can shape how people engage with visualized data.
- The paper asks how visualizations can attract attention without compromising data integrity, particularly among underrepresented populations with diverse educational and socioeconomic profiles.
Why might rural populations be different?
Rural settings combine technological, educational, and geographic constraints that can limit access to visualization. The study therefore examines initial perceptions among a diverse group of central Pennsylvania community members.
- Economic and infrastructure obstacles in rural America can undermine the impact of visualization tools, with lower-income people more reliant on phones or public computers.
- Rural environments introduce additional constraints to visualization literacy, including financial barriers, geographic isolation, and limited adult-education coverage.
- The study focuses on rural populations because they are underrepresented in visualization literature despite comprising nearly 60 million people in the United States.
- Researchers interviewed 42 participants from a university, construction site, and farmers market in central Pennsylvania.
- Ten color visualizations about drug impacts in the United States were selected to vary in form, visual appeal, and source.
- Participants reported varied ages, education, political affiliations, graph familiarity, personal drug-related impact, and family incomes.
Procedure
Participants ranked and discussed ten visualizations in semi-structured interviews, after which researchers revealed sources and collected demographic information. Analysis used iterative grounded-theory-informed coding of all 42 transcripts.
- Participants received as much time as needed to rank ten graphs from most to least useful, using a prompt designed to elicit their own values.
- Follow-up questions addressed differing rankings of line graphs, maps, and infographics among subsets of participants.
- Researchers revealed chart sources, provided brief context for unfamiliar sources, and asked whether participants would revise rankings or rationales.
- Demographic questions followed the interview to prevent priming, and interviews typically lasted approximately 15 minutes with $10 compensation.
- Researchers generated codes and themes from transcripts rather than imposing existing visualization theories, then iteratively revised them across the 42 transcripts.
- Aggregated code tallies and study materials were deposited in an open repository, while code frequencies require caution because interviews were semi-structured.
Rankings Overview: Clarity, Simplicity, and Color
Participants’ rankings reflected diverse views, but clarity, simplicity, color, and visual appeal were recurring considerations. Infographics produced especially polarized responses, and perceived clarity often meant quickly grasping a chart’s gist rather than deeply understanding its data.
- Colorful and confusing were the most common graph codes, each occurring 29 times, followed by clear and simple at 26 times each.
- Participants broadly favored straightforward visual encodings, while their diverse rankings could be obscured by aggregated summaries.
- Participants associated simplicity with fewer lines and less detail, which some believed would make graphs understandable to more people.
- Perceived clarity often meant quickly extracting a visualization’s gist rather than deeply understanding the underlying data.
- Sixteen participants identified color as distinguishing Graph I from Graph G, although they were unclear whether it improved appeal or visual encoding.
Data is Personal
Participants’ responses show that personal experience with substance use and addiction could outweigh visualization style or clarity when shaping attention and rankings. Relatability made drug-related content feel urgent and personally meaningful.
- Personal connections to alcohol made Graph J valuable to participants despite differences in age and educational background.Participants linked the chart to their own drinking or to alcoholism affecting someone important in their lives.
- The findings suggest that personal relevance can supersede design characteristics such as style, clarity, or ease of understanding.More than 20 instances involved participants referencing a relatable component of a graph’s content.
- Opioid-related experiences shaped both preferences for opioid graphs and criticism of graphs that omitted opioid information.Participants referenced local overdose patterns, acquaintances with opioid problems, and the perceived importance of opioids as a cause of death.
- 22 of 42 participants reported being personally impacted by drug abuse at level 5 or higher, including 8 who selected 7 out of 7.These experiences were often unspoken during interviews, but may have strongly influenced perceptions.
Geographical Information: Where am I in the data?
Geographic framing produced diverse reactions, but participants repeatedly attended to places connected to their own lives. Local and national scope therefore shaped perceived relevance, while interactive behavior remained outside the study’s direct evidence.
- Participants described maps as clear or simple in some cases and confusing or cluttered in others.Across two otherwise similar maps, clarity was coded 8 times, simplicity 6 times, confusion 5 times, and clutter 5 times.
- Six participants focused on locations where they currently live or had lived previously when discussing geographic data.Comments referred to places such as West Virginia, Pennsylvania, and participants’ own states.
- Maps in the study were designed to be interactive, so reactions to static images may not translate to interactive versions.The authors explicitly limit this interpretation to perceptions of the static presentations.
- A participant ranked a low-performing county heat map highly because it included their own county and felt personally relatable.The paper notes that participants often focused on their local region even though national trends may use larger samples.
Social Framing: Will this help other people?
Some participants evaluated visualizations not only for personal usefulness but also for how they might serve other people. Judgments of infographics varied with perceived credibility, audience engagement, familiarity, and statistical comfort.
- Some participants considered infographics less serious than conventional charts, even when they found them attention-getting.One participant said their novelty helped attract attention but reduced their ranking because they seemed less credible.
- A school principal ranked infographics according to what would engage students and parents, combining informational content with visual appeal.This reflects an outward-looking evaluation of charts for a broader audience rather than only for personal use.
- Infographics were divisive: Graph J received the most polarizing rankings of any chart.Some participants found infographics clear, simple, and attractive, while others described them as boring, bizarre, childlike, or overly decorative.
- One older participant reported needing to study newer visualizations such as infographics more carefully, although this was not a consistent age trend.The authors connect this observation cautiously to prior work on older audiences and visualization novices.
- Different reactions to Graph F’s clarity may have reflected statistical familiarity associated with education or numerical work experience.Three of four participants who found the graph simple had college degrees, while the fourth frequently worked with numerical data.
Why don’t people change their rankings?
Many participants kept their initial rankings after learning chart sources because they viewed source as irrelevant, trusted all sources equally, or prioritized unchanged visual criteria. Education was associated with changing rankings, while the authors caution that first exposure and anchoring may shape responses.
- 25 of 42 participants kept their initial rankings after chart sources were revealed.Among 22 participants whose rationales were categorized, source irrelevance, other criteria, no reason, and equal trust were reported.
- 52% of participants who maintained their rankings had no post-secondary education, compared with 1 of 17 participants who changed rankings.The authors hesitate to draw conclusions from the sample but note implications for generalizing studies of educated participants.
- 12 of 22 participants who did not change rankings said the source was irrelevant or that all sources were equally trustworthy.Some participants characterized information as objective regardless of its origin.
- Five of 22 participants expressed reluctance to change their rankings regardless of new information.Other participants maintained rankings because their original criteria, such as readability, had not changed.
- The authors suggest that anchoring may explain why initial rankings persisted even when participants acknowledged the importance of data sources.They caution that final rankings might differ if source information had been provided before the first judgment.
Why do people change their rankings?
Participants changed rankings after source information was revealed, often giving greater weight to source validity, institutional affiliation, and local relevance. Among the 17 participants who changed rankings, academic and government sources commonly received higher evaluations, while personal associations also shaped responses.
- 17 of 42 participants changed their rankings after the visualization sources were revealed, placing greater emphasis on source validity.
- 8 out of 17 participants ranked the NIDA, Drexel, and NVSS graphs higher after their academic or government sources were revealed.These sources had initially been criticized for confusing, childlike, or lacking-credibility designs.
- Figure 6 orders charts by mean improvement in rankings and shows the distribution of ranking shifts among the 17 participants who changed rankings.Positive shifts indicate improved rankings.
- Drexel’s graph received a positive ranking alteration from 8 participants despite initially being perceived as the most confusing chart.Its academic status or familiarity as a Pennsylvania institution may have contributed, although the study could not distinguish between them.
- One participant preferred a local newspaper over The New York Times because local proximity made the source more personally relevant.The authors caution that this individual perspective cannot support generalization, but suggest locality merits further study.
Trust and Political Identity
Trust in data sources varied substantially across participants and sometimes aligned with political identity. Political affiliation was especially associated with ranking changes for The New York Times, AGRiMED, and Breitbart, suggesting that institutional reputation did not uniformly determine trust.
- Participants’ trust in governmental sources varied significantly, with contrasting liberal and conservative views of the CDC.One participant expressed unconditional trust in the CDC and NIH, while another believed the CDC hid information.
- Rank changes aligned with political identity for The New York Times, AGRiMED, and Breitbart.Participants disclosed political affiliation only after completing the interview.
- Liberal participants were more likely to lower rankings for Breitbart and AGRiMED visualizations.Breitbart is described as a far-right news outlet, while AGRiMED is a licensed medical cannabis cultivation company.
- Political biases and beliefs may reduce the impact of highly regarded visualization institutions, including The New York Times.
5 DISCUSSION
The interviews show that personal experience, education, visual preferences, political identity, and source perceptions jointly shaped how participants prioritized visualizations. These findings highlight the importance—and difficulty—of making personal relevance visible early while recognizing limits on generalization and risks of misinformation.
- Discussion: The 42 interviews identified clarity, simplicity, color, visual appeal, source, education, political identity, and personal experience as influences on visualization priorities.Personal experience was the dominant theme selected for further discussion.
- Visualizations are Personal: Participants affected by abuse or addiction gravitated toward graphs representing those substances, while geographic graphs were judged partly by how easily participants found their home state.
- Visualizations are Personal: Personal framings may alter attention before interaction, so visualizations may need to foreground personal dimensions at first encounter.The authors connect this challenge to designs that support personal exploration and goals.
- Visualizations are Personal: Reluctance to change rankings and perceptions that information is objective suggest that first exposure to a visualization may be critical.The paper raises the design challenge of ensuring that initial visualizations are reliable and can alter prior impressions.
- Limitations: Because education and political identity aligned with participant priorities, findings from studies focused on highly educated students may not generalize to the broader public.
- Limitations: The study involved 42 participants reflecting on 10 graphs, and its sample may not represent hard-to-access marginalized communities in the United States.Political affiliations were more liberal than voting records would suggest.
- Limitations: The authors call for replication with controlled studies, different populations, larger samples, and recruitment strategies targeting difficult-to-access groups.
- Broader Impacts: Designing communication for underrepresented groups may also create opportunities for misinformation, motivating information systems that combat rather than amplify it.