Consequential insights for advancing informal STEM learning and outcomes for students from historically marginalized communities

The paper addresses persistent inequities in STEM education for students from historically marginalized communities (HMC) and examines whether informal, research-based STEM experiences can equitably support STEM literacy and sustained interest. SEE (Systems Education Experiences) at the Institute for Systems Biology expanded from a 320-hour internship to four models (320-h internship, 90-h course, 40+ hour workgroup, 22-h short courses) to scale participation and better fit student needs. The study focuses on three questions: (1) what progress on STEM pathways do students from HMC make through informal STEM experiences; (2) what successes and challenges emerge when HMC youth engage in authentic STEM; and (3) what aspects of informal experiences support progress on STEM pathways. The context underscores the importance of STEM literacy for societal challenges (climate, health) and the underrepresentation of HMC in STEM, motivating investigation of which program features, contexts, and durations most effectively foster equitable outcomes.

The theoretical framework integrates sociocultural perspectives of learning as transformation of participation within communities of practice and acknowledges how learning settings can privilege dominant discourses. Drawing on frameworks such as the Equity-Oriented Conceptual Framework for K–12 STEM Literacy and the concept of student thriving, the study centers six overlapping constructs: engagement; awareness and intent (STEM pathway progression); identity; social capital; and 21st Century Learning Skills. Prior literature suggests consequential participation broadens access when youth engage in authentic cross-cutting STEM practices (e.g., modeling, computational and systems thinking), develop science-linked identities, and connect STEM to everyday life and community advocacy. Informal settings (museums, research institutes) play a critical role in the broader STEM ecosystem, yet there is a need to pinpoint which aspects of programs most effectively support HMC learners. The paper embeds a literature review under each construct, citing work on engagement’s relationship to identity and relevance; early formation of STEM pathways and thriving beyond graduation metrics; identity as socially negotiated and context-dependent; social capital via developmental relationships; and 21st Century Skills overlapping with disciplinary practices. These guide the selection of constructs and informed the development/adaptation of measurement tools.

Design: Mixed-methods study analyzing retrospective alumni survey data (2003–2019) and retrospective pre–post participant survey and open-ended responses (2019–2021), aligned to six STEM equity constructs. Participants: SEE has served 800+ students; 415 completed immersive experiences at ISB (80% from HMC). The study analyzes data from 239 students (58% of 2003–2021 participants): 56 alumni (2003–2019; internships or ambassadorships) and 183 participants (2019–2021) across four experience models: 320-h internship (2019 in-person; 2020–2021 remote), 90-h course (2019 in-person), 40+ hour workgroup (2020 remote), and 22-h short courses (2021 remote, with 2-h Tier 1 + 20-h Tier 2). Recruitment/Selection: Common application included prompts, resume, transcript, and recommendation. SEE staff reviewed, interviewed, and selected using a rubric considering background, social environment, and prior STEM exposure. Priority was given to HMC students and those with limited STEM experience. Instruments and Data Collection: Two online surveys (SurveyMonkey) drew from validated tools; Likert 4-point scale (Strongly Agree=4 to Strongly Disagree=1). Alumni survey asked: extent SEE, high school, and other extracurricular experiences supported five STEM pathway components (course of study, career pathways, content/practices, barriers, societal connections). Participant survey was retrospective pre–post; items mapped to constructs with counts per construct (Engagement post-only: 7 items; Awareness pre–post: 3; Identity post-only and pre–post: 3 each; Intent pre–post: 2; Social capital post-only and pre–post: 2 and 3; 21st Century Skills pre–post: 4). Open-ended questions (2020–2021) elicited which activities contributed to growth in each construct. For 22-h courses, Tier 2 survey captured three timepoints; analysis reported endpoints (before Tier 1 vs after Tier 2) for alignment with other models. Analysis: For Likert items, ordinal values were averaged per construct for each student; only complete pre–post pairs were analyzed within constructs. Means, percent change, descriptive stats, and paired t-tests (two-sided; significance thresholds noted in figures/tables) were computed; violin and scatter plots visualized distributions. Alumni comparisons used means and t-tests between SEE vs high school and SEE vs other extracurriculars. Qualitative responses were open coded by the team to produce themes, with iterative refinement and triangulation. Validity and Reliability: Instruments adapted from validated tools; items tailored to program context to ensure ecological validity. Acknowledged potential positive-response bias (mitigated by framing, item order/phrasing). Demographic disaggregation was limited to avoid identifiability in small cohorts. Tier 1 to Tier 2 responses were not longitudinally linked due to anonymization.

• Alumni pathway outcomes and comparative impact: Among 56 alumni (37% response rate; 62% HMC), 96% took advanced STEM in high school; 98% went directly to undergraduate; 78% majored/minored in STEM; 84% pursued or planned a medical/graduate/professional degree; primary fields: research (58%) and healthcare (38%). Alumni rated SEE significantly higher than high school across all five pathway components; largest differences: career pathways (mean 3.4 vs 2.4) and barriers (3.0 vs 2.0). Compared with other extracurriculars, SEE was significantly higher in content/practices (3.5 vs 2.9) and societal connections (3.0 vs 2.5).
• Engagement: Very high satisfaction and interest across experiences (means ~94% and 96% positive). In the 90-h course, clarity of program goals had the lowest positive rate (69%), but affected students still showed gains in other constructs.
• Awareness: All experiences increased awareness of STEM careers and pathways. Lower percent change for the 22-h short courses (≈11%) reflected higher pre-awareness (pre mean 3.1) relative to other models (pre 2.4–2.6); post means converged (3.4–3.6).
• Identity: Post-only identity environment items were ≥83% positive; pre–post identity increased in all models with overall percent changes of ~17.7% (320-h), 19.2% (90-h), 12.7% (40+ hour), and 7.9% (22-h), with smaller changes due to high pre-identity (pre means 3.2–3.4; post 3.6–3.9).
• Intent: High pre-intent (pre means 3.4–3.6) and near-ceiling post values; 100% of 320-h and 40+ hour students already agreed they would consider a STEM career pre; 90-h and 22-h cohorts were 94% and 98% pre, respectively; post values were ~100% across models.
• Social capital: Significant gains in two of three social capital items: speaking with a STEM professional about their job and knowing someone outside school to learn more about STEM. Family/friend encouragement was high at pre and post; smaller change. Plans to stay connected (post-only): mentors—100% (320-h), 100% (90-h), 88% (40+ hour), 83% (22-h); peers—100%, 94%, 85%, 56%, respectively.
• 21st Century Learning Skills: Gains across models in adapting when plans change and managing time independently; consistently high ability to work with different people; somewhat lower post confidence in risk-taking in the 90-h (88%) and 22-h (89%) cohorts relative to other skills.
• Remote vs in-person (320-h internships): Comparable outcomes across 2019 (in-person) and 2020–2021 (remote) cohorts; significant pre–post gains within cohorts. Pre, post (percent change): Awareness—2019 2.7→3.7 (38%), 2020 2.5→3.7 (48%), 2021 2.3→3.4 (49%); Identity—2019 3.5→3.8 (8%), 2020 3.3→4.0 (20%), 2021 2.9→3.8 (33%); Intent—2019 3.7→4.0 (10%), 2020 3.7→4.0 (8%), 2021 3.1→4.0 (29%); Social Capital—2019 3.0→3.8 (26%), 2020 3.0→3.8 (28%), 2021 2.9→3.9 (34%); 21st Century Skills—2019 3.3→3.7 (11%), 2020 2.9→3.8 (29%), 2021 2.7→3.3 (22%). Differences in post 21st-century skills favored 2019/2020 over 2021.
• Activity features linked to outcomes (open-response coding): Identity and intent were supported by all identified activity themes. Highest attributions: Using scientific tools/materials (identity 26%, intent 17%); guest speakers/interviews/career videos (awareness 72%; intent 29%; social capital 20%; identity 9%; 21st skills 5%); research activities (identity 17%; intent 17%); collaboration and group work (21st skills 47%; social capital 22%). Presentations/project work supported social capital (24%), 21st skills (14%), identity (10%), awareness (7%), and intent (7%).
• Trade-offs in shorter duration models: Some 22-h participants did not report gains in understanding possible STEM careers or day-to-day STEM work; also lower increases in talking with professionals and knowing external STEM helpers relative to longer models.

Findings directly address the research questions. Q1: HMC students progressed on STEM pathways via informal experiences, with alumni outcomes and ratings showing SEE uniquely bolstered informed decisions about study, understanding STEM careers, authentic content/practices, and societal connections, exceeding high school and other extracurricular programs. Q2: Successes included consistently high engagement, gains in awareness, identity, intent, social capital, and 21st-century skills across varied durations and modes. Challenges emerged in short-duration offerings (22-h) where limited time constrained depth of career-connected learning and peer network building; social capital with peers scaled with time-on-task and cohort size. Remote programming can be as effective as in-person when designed with strong mentoring and authentic tasks, though logistical barriers (e.g., at-home labs) can hinder some activities. There remains a need across settings to better support students in identifying and navigating barriers in STEM and in making explicit societal connections—areas rated lower overall by alumni even when SEE outperformed comparisons. Q3: Aspects that most supported progress were authentic research participation, use of professional tools, direct and diverse interactions with STEM professionals, structured group work, and culminating public presentations. Intentional structures—regular mentor interactions, near-peer support, explicit goal alignment, and visible recognition of student contributions—facilitated identity development, social capital, and persistence. Overall, the study supports designing co-created, principle-driven informal STEM experiences that can scale with minimal resources while maintaining equitable impact for HMC students.

Informal, research-based STEM programs can equitably advance STEM literacy and pathway progress for high school students from historically marginalized communities across a range of durations (22–320 hours) and delivery modes (in-person and remote). SEE experiences uniquely supported informed academic decisions, awareness of career pathways, mastery of authentic content and practices, and recognition of STEM’s societal relevance. Key components include authentic research tasks, access to professional tools, multiple interactions with diverse STEM professionals, collaborative projects, and public sharing of work. Practical recommendations include: co-creation with youth and educators; explicit, ongoing goal-setting and progress tracking; integrating near-peer mentors to strengthen cohort connections; providing structured, career-connected activities even in short courses; and scaffolding to help students recognize and navigate barriers and connect STEM to societal issues. These strategies enable scalable, resource-efficient programs that can broaden participation in STEM, enhance real-world problem solving, and better prepare society to benefit from scientific advances. Future work should continue refining supports for barrier navigation and societal connection-making, and optimize remote/hybrid designs to mitigate logistical challenges while preserving authentic practice and mentoring.

• Response and measurement bias: Retrospective pre–post design may inflate gains; potential positive-response bias as surveys were administered by program staff (mitigated via framing and item design).
• Anonymity constraints: For 22-h short courses, Tier 1 and Tier 2 surveys were not linked longitudinally; demographic disaggregation was limited to avoid identifiability in small cohorts.
• Sample size variability: Some cohorts and item-level Ns were small, affecting stability (e.g., single responses influenced percentage shifts in 21st-century skills items).
• Generalizability: One organization (ISB/SEE); content and contexts varied across models and years; lack of case–control designs for some comparisons (e.g., remote vs in-person across all models).
• Implementation constraints: Remote hands-on labs faced logistics (material delivery, home space) that may have dampened some outcomes; short-duration (22-h) models had limited time for deep career exploration and peer network development.