How to Do Real Behavioral and Social Sciences Research at Home: A High School Student’s Guide to Free Tools, Affordable Kits, and Public Databases

Behavioral and social science research used to mean a university lab, a recruited adult sample, and a stack of paper questionnaires. Now a laptop, a phone, and a free survey link can do most of it.

This guide is your starting point. It covers three things: the small kit you can put together for under the cost of a textbook, the free software professional researchers actually use, and the public datasets that hold decades of human behavior waiting to be re-analyzed.

Why this is possible now

Three shifts opened this field to high school students.

First, experiment builders went free. PsychoPy, jsPsych, and lab.js let you build the same reaction-time, memory, and decision tasks that appear in peer-reviewed journals, then host them in a browser so anyone with a link can participate.

Second, consumer sensors caught up to research-grade hardware for many questions. A $90 chest strap reads heart-rate variability cleanly. A webcam plus open-source computer vision tracks gaze, posture, and micro-expressions. A smartwatch exports sleep stages by the night.

Third, the world's social data went public. Decades of surveys, police stops, language corpora, and platform archives sit on open repositories. You can ask a real question of real data on a Tuesday afternoon.

Put together, a kitchen counter plus a laptop can now run a preregistered experiment, collect physiological signals, and merge the results with a national dataset by the end of a weekend.

The behavioral and social sciences home kit

Group your kit by what each piece lets you measure.

Devices you probably already own

• A laptop with a webcam (for eye-tracking, pose, and online experiments).
• A smartphone (for experience-sampling apps, audio recording, and survey delivery).
• A smartwatch or fitness tracker you can export data from (Fitbit and Apple Watch both allow this).

Physiological add-ons (optional, under $300 total)

• A consumer HRV chest strap such as the Polar H10 (~$90) for clean heart-rate variability.
• A Muse 2 or Muse S EEG headband (~$250) for frontal alpha and basic neurofeedback work.
• An Arduino with a GSR or pulse sensor (~$30) if you want to build your own physiology rig.
• A Raspberry Pi (~$40) for any always-on data collection station.

Survey and recruitment tools

• A free Google Forms or Qualtrics free-tier account for consent and questionnaires.
• A small Prolific budget (often under $50) if you need participants outside your school.
• A signed parental-consent template and a one-page protocol for your school or fair IRB.

Behavioral coding and recording

• OBS Studio plus your webcam for recording sessions.
• A quiet room and a tripod or stack of books to keep the camera still.
• A printed observation sheet for in-person studies before you move to BORIS for formal coding.

A research notebook

• A bound notebook or a dated digital doc for every session, every change, and every decision.

Total cost range for a strong starter setup: roughly $0 to $400, depending on how much physiology you want to record.

Signature technique: browser-based experiments with webcam sensing

The one technique that unlocks the most projects in this field is the in-browser experiment. You build a task once, share a link, and collect data from anywhere. Adding webcam-based eye-tracking or pose detection turns it into something close to a lab study.

Here is the five-step workflow:

1. Write your one-line question. Example: does font style change retention after a one-week delay? The question fixes your design before you touch code.
2. Build the task in jsPsych or lab.js. Both are free JavaScript libraries with templates for reaction time, memory, Stroop, n-back, and survey blocks. Copy a template, swap in your stimuli.
3. Add webcam sensing if it helps. WebGazer.js gives rough gaze position. MediaPipe gives face landmarks and pose. OpenCV in Python can post-process your recordings for pupil size or micro-expressions.
4. Host the experiment. Pavlovia and Gorilla host PsychoPy and jsPsych studies for low cost. You can also self-host a static jsPsych build on GitHub Pages for free.
5. Preregister, then collect. Write your hypothesis, sample size, and analysis plan on the Open Science Framework before you launch. Recruit through school clubs with consent, or use a small Prolific batch for an outside sample.

This same workflow scales from a 1-week pilot to a full ISEF project.

The dry-lab side: free software you can install today

Group your software by what it lets you do.

Experiment builders

• PsychoPy: desktop and online experiments with precise timing.
• jsPsych: JavaScript library for browser-based cognitive tasks.
• lab.js: drag-and-drop builder that exports to a browser.
• OpenSesame: Python-based builder favored for psychophysics.

Statistics and modeling

• JASP and jamovi: free point-and-click stats with Bayesian options built in.
• R with lme4, brms, and lavaan: mixed models, Bayesian regression, structural equation modeling.
• Python with pandas, statsmodels, and scikit-learn: the workhorse stack for any data analysis.
• PyMC: hierarchical Bayesian modeling, useful for reinforcement-learning fits.

Behavior and signal analysis

• BORIS: frame-by-frame coding of video for observational studies.
• OpenCV and MediaPipe: face, gaze, hand, and pose tracking from any webcam.
• DeepLabCut: markerless pose estimation when MediaPipe is not enough.
• Praat: acoustic analysis of speech.
• ELAN: time-aligned annotation of audio and video.
• pyannote.audio: speaker diarization for classroom or group recordings.

Text and language

• spaCy and NLTK: tokenization, parts of speech, named entities.
• Hugging Face Transformers: ready-to-use models for sentiment, embeddings, and topic work.
• BERTopic: topic modeling on documents and social media.
• Whisper: high-quality speech-to-text from audio.

Simulation and networks

• Mesa: agent-based modeling in Python.
• NetworkX: graph analysis of social and citation networks.

Running these tools yourself changes how research feels. You stop reading papers as finished objects and start reading them as code you could rerun tonight.

Public databases that count as real data

Group your data sources by what they measure.

Survey and panel data

• ICPSR: thousands of social science studies with documentation.
• General Social Survey (GSS): US attitudes since 1972.
• American National Election Studies (ANES): political behavior over decades.
• World Values Survey: cross-national attitudes and beliefs.
• Pew Research datasets: clean topical surveys with methodology files.
• PISA well-being supplements: cross-national adolescent data.

Government and economic

• US Census and ACS: demographics at neighborhood resolution.
• Bureau of Labor Statistics (BLS): CPS microdata for labor questions.
• CDC BRFSS: state-level health and wellbeing.
• OECD and Eurostat: cross-country policy and outcome data.
• UN Data: global indicators on development and equity.

Behavior and platform archives

• Reddit Pushshift archives: long-running conversation data.
• Twitter/X academic archives and Common Crawl: large public text corpora.
• Google Trends and Wikipedia pageviews: free signals of public attention.
• Stanford Open Policing: traffic stops across US jurisdictions.
• FBI UCR and city open-data portals (Chicago, NYC): crime, 311 calls, transit.

Language and development

• CHILDES: child language transcripts across many languages.
• COCA and COHA: contemporary and historical American English corpora.

Psychology-specific

• OpenPsychometrics: large item-response datasets on personality and ability.
• Many Labs replication data: multi-site replications with documentation.
• Project Implicit: implicit-association data across years.
• OpenNeuro and Neurosynth: open brain-imaging data and meta-analytic maps for purely computational projects.

Re-analyzing public data is a legitimate research path on its own. A careful difference-in-differences study on Open Policing data is real research, full stop.

How to combine wet and dry: the strongest project shape

The strongest behavioral projects fuse hands-on measurement with computational analysis.

Pattern A: small primary study plus public benchmark. You run a focused experiment with peers (say 30 to 80 participants), then compare your effect size to a public meta-analysis or replication dataset. This anchors a tiny sample to a much larger conversation.

Pattern B: public dataset plus a small validation study. You start with a pattern you find in ICPSR, GSS, or a platform archive, then collect a tiny new sample that tests one specific prediction the pattern makes. The public data gives scope, your study gives causal traction.

Judges respond to this shape because it shows you can both collect clean data and reason about data you did not collect.

Choosing a phenomenon that has not been done

Novelty in behavioral science is rarely about a brand-new question. It is about a new combination of measure, sample, and design. Use a three-step check.

1. Google Scholar search. Take three or four keywords from your one-line question and look at the most-cited recent reviews. Note what their "future directions" sections say.
2. Society for Science abstracts archive. Read past ISEF and Regeneron abstracts in Behavioral and Social Sciences. You are not looking for forbidden topics. You are looking for which measures, samples, and designs have already been combined.
3. PubMed and PsycInfo (free abstract search). Run your keywords as a title search. Read the abstracts of the five most recent papers. Adjust your question so that at least one of measure, sample, design, or context is new.

Finding adjacent prior work is good news, not bad news. It means your question is in a live conversation other researchers are already having.

A realistic timeline

• 1 to 2 weeks: a focused replication or measurement pilot. Run a known task on 20 or 30 peers and compare your effect to the published estimate.
• 1 to 2 months: a full hybrid project for a regional fair. One primary study plus a public-data analysis, preregistered, with a clean writeup.
• Full year: an ISEF-track project. Multiple studies, a preregistered confirmatory phase, ethics review, and a long literature review.

First-time researchers should start with the 1 to 2 week version.

A starter checklist

1. A clean, quiet workspace with a webcam, a tripod or book stack, and a closing door.
2. A free Google Colab account and a local Python environment with pandas, scikit-learn, and statsmodels installed.
3. JASP or jamovi installed for fast statistical checks.
4. Your experiment builder of choice (PsychoPy, jsPsych, or lab.js) installed and running a demo task.
5. A signed parental-consent template and your school or fair's IRB-equivalent form ready to submit.
6. A bound or digital lab notebook with the date on every entry.
7. A written one-line research question taped where you can see it.

When all seven are in place, you are ready to pick a phenomenon.

Where to go next

ISEF places behavioral and social sciences work in these subcategories. Pick the one that interests you most.

• Behavioral Neuroscience (NEU): how brain and body signals (HRV, EEG, pupil, sleep) shape attention, stress, learning, and decision-making.
• Development (DEV): how cognition, language, and behavior change with age, from early childhood through adolescence.
• Cognitive Psychology (COG): memory, attention, problem-solving, reasoning, learning, and the design of tasks that probe them.
• Social Psychology (SOC): persuasion, conformity, group dynamics, identity, prosocial behavior, and online social influence.
• Sociology and Anthropology: population-level patterns in policy, inequality, culture, networks, and language using public data.
• Other (OTH): cross-disciplinary work, behavioral economics, human-AI interaction, accessibility, and applied mental-health tools.

Each subcategory has its own MehtA+ project guide built on the kit and software on this page. Pick the one that pulls at you and start there. The field that used to require a lab now fits on your desk.

Project ideas in this category (66)