Boost Exhibit Design: AI Qualitative Analysis for Museums

Evidano is an AI-powered qualitative data analysis platform that helps researchers analyze interviews, open-ended surveys, and documents. This post translates a PLOS ONE observational study of the National Museum Prague Children’s Museum into practical steps for researchers and exhibit teams using AI qualitative analysis for museums. According to PLOS ONE, the study observed 720 interactions involving 2, 173 visitors during a 2.5 month field period (16 March to 30 May 2024), and found that the intended 6–10 age cohort made up only 23% of child visitors while 72% were younger children, suggesting a mismatch between constructivist exhibit design and actual audiences PLOS ONE. Below you will find extractable findings, a data snapshot, implications for museum researchers, example AI workflows, and a short FAQ that together show how AI-enabled qualitative research can turn observational notes into design decisions.

Key Takeaways

The PLOS ONE study observed 720 interaction records covering 2, 173 visitors from 16 March to 30 May 2024 and concluded the constructivist Children’s Museum attracted predominantly younger children rather than the intended 6–10 cohort, with the authors noting "the primary target group constituted only 23% of the child visitors, with 72% being younger children" (PLOS ONE).

720 observations and 2, 173 visitors were recorded during the 16 March to 30 May 2024 field period, according to PLOS ONE.
PLOS ONE reported that the exhibition’s target age group (6–10) made up only 23% of child visitors while 72% were younger children, in data collected through structured observation.
PLOS ONE found engagement varied by unit: average dwell time ranged from about 2.5 minutes at one unit (U2) to 14 minutes at others (U1 and U5) across the same observation period.
PLOS ONE recorded markedly low facilitator visible engagement overall (mean = 0.27 on a 0–3 scale), with higher facilitator engagement specifically in U5.

What happened and how the study measured it

Answer: PLOS ONE used a structured, protocol-driven observational method to quantify visitor interactions and engagement across five exhibition units between 16 March and 30 May 2024.

According to PLOS ONE, 19 trained observers collected 720 unit-level observations representing 2, 173 individuals, using a 4-point Likert coding scheme for interaction dimensions (e.g., communication, creating, focus) and for engagement of children, adults, and facilitators.

According to PLOS ONE, observers combined scalar protocol data with open-ended field notes and schematic area maps, and then analyzed interaction dimensions with ordinal regression models to test effects of unit, age composition, slot, capacity, group size, and day of week.

According to PLOS ONE, the study treated the visiting family as the analytical unit for interactions, excluded organized school groups, and made datasets and R scripts available in the study OSF repository OSF repository.

Findings snapshot

Date / Period	Metric	Value (as reported)	Implication
16 March–30 May 2024	Observations	720 unit-level observations	Large structured sample for unit-level interaction comparisons
16 March–30 May 2024	Individual visitors observed	2, 173 visitors (reduced N = 1, 759 for unique-individual composition analysis)	Enables demographic profiling and group-composition analysis
30 June 2023	Exhibition opening date	ChM opened 30 June 2023	Context: new constructivist concept still establishing its audience
Observation period	Target age share	23% were children aged 6–10; 72% were younger children (PLOS ONE)	Design/marketing mismatch: exhibition draws younger cohort than intended
Across units	Dwell time	Mean per unit from ~2.5 to 14 minutes	Unit design strongly affects time-on-task and engagement

Implications for museum researchers and exhibit teams

Answer: The PLOS ONE results imply researchers should combine structured observation with targeted audience segmentation and rapid thematic synthesis to close the design–audience gap.

According to PLOS ONE, a constructivist exhibition that relies on adult facilitation may underperform when the visiting population skews younger than expected; designers should re-evaluate visitor composition and slot/capacity policies to match affordances to users.

According to PLOS ONE, unit-level differences (for example U2’s short dwell time vs U5’s long dwell time) indicate that small design changes per unit can yield measurable changes in focus, creativity, and communication.

Actionable decision: PLOS ONE recommends adjusting marketing to attract older children and experimenting with alternative entry sequences (for example starting visits at U5) to improve thematic coherence and reduce cognitive overload.

How Evidano helps museum researchers translate observations into decisions

Problem: Large, messy observational datasets slow synthesis

Solution: Evidano ingests transcripts, observation protocols, and field notes, and auto-extracts themes, frequencies, and cross-segment comparisons.

Context: PLOS ONE collected 720 protocol rows and rich field notes over 2.5 months; Evidano can map scalar protocol fields to visualizations (co-occurrence networks, hierarchical code trees) to surface which units drive creativity or low engagement.

Problem: Identifying age-group patterns across units is time-consuming

Solution: Evidano supports cross-segment analysis by age, slot, and unit so teams can reproduce the PLOS ONE contrasts (for example, comparing dwell time and creativity scores between U1 and U3) in minutes rather than weeks.

Note: PLOS ONE reported that older children (6–10) showed higher engagement in certain units; Evidano’s segment filters let you test similar hypotheses on your own dataset and produce exportable tables and charts.

Problem: Turning qualitative notes into testable design changes

Solution: Evidano creates thematic summaries and recommended taggings from observer field notes, converting statements like "the interactive wall may not connect visitors thematically" (PLOS ONE) into prioritized tickets for designers and facilitators.

Evidano also integrates with transcription and translation features (see Evidano features) to harmonize multi-language field notes and create a single searchable corpus.

Problem: Need for rapid facilitator training materials

Solution: Evidano can synthesize common visitor utterances and successful facilitator prompts, producing short scripts and question sets that align with the study’s recommendation to provide facilitators with "opening lines and sets of questions" (PLOS ONE).

For organizations that require secure handling, Evidano provides encryption and guarantees data is not used to train third-party models; see the platform overview at Evidano features.

FAQ: AI qualitative analysis for museums

How can AI help analyze structured observation protocols like those in the PLOS ONE study?

Direct answer: AI can automate coding, theme extraction, and cross-segment frequency counts so teams can move from raw protocols to prioritized findings quickly.

Supporting detail: The PLOS ONE study used 4-point Likert items across multiple interaction dimensions; AI-enabled coding can ingest those scalar fields alongside free-text field notes to produce combined quantitative+qualitative summaries and visualizations.

Can AI detect which exhibit units produce the most creativity or focus?

Direct answer: Yes, AI-enabled cross-tabulation and regression-ready exports can surface unit-level differences that mirror the PLOS ONE ordinal models.

Supporting detail: PLOS ONE reported U3–U5 had higher creativity scores and U2 had shortest dwell times; an AI pipeline can reproduce these contrasts, flag high-effect units, and create exportable charts for stakeholder review.

Is it ethical to use AI on observations that include children?

Direct answer: Yes, when you follow privacy, consent, and data minimization practices appropriate to research with minors.

Supporting detail: PLOS ONE described non-participant overt observation without collection of personal identifiers; for AI workflows, remove or redact PII at ingestion, follow institutional review guidance, and use secure platforms that do not train external models.

How fast can a museum team go from raw observations to prioritized design changes using AI?

Direct answer: A single analyst can often produce actionable insights in 24–72 hours using AI-assisted thematic coding and cross-segment comparisons.

Supporting detail: The PLOS ONE dataset of 720 observations would typically require weeks to synthesize manually; AI-assisted pipelines accelerate coding, let teams iterate on codes, and generate visualizations for stakeholder meetings within a few days.

Conclusion & Next Steps

The PLOS ONE study of the National Museum Prague Children’s Museum demonstrates how structured observation plus qualitative field notes reveal unit-level strengths and a pronounced audience/design mismatch (for example, 23% of child visitors were the intended 6–10 cohort while 72% were younger children according to PLOS ONE).

Applying AI-enabled qualitative research tools lets museum researchers reproduce those unit contrasts, prioritize low-effort high-impact design fixes, and generate facilitator scripts that the study recommends.

If you want to test this workflow on your own observation protocols and field notes, you can start a project and import spreadsheets, transcripts, or observer notes into Evidano now.

Get hands-on: Try Evidano for free.

Topics

AI qualitative analysis for museums
qualitative analysis museums
museum visitor research
AI-enabled qualitative research
exhibit evaluation

Keep reading

Browse all articles