Home/Research/ARforEarth Research Foundations guide
RESEARCH FOUNDATIONS / EDITION 1
A careful question. A credible contribution.
Read the complete course guide before the Research Foundations assessment. Completing the course and passing the assessment unlocks new research projects in your account.
Edition 1 · November 2026 · Use your browser’s Print command to save a PDF.
Read first. Answer independently.
Complete all 8 required lessons before starting. The assessment contains 12 randomized questions, allows 30 minutes and requires at least 10 correct answers to pass the 80% threshold. You may start up to three attempts in a rolling 24-hour period. Once an attempt starts, its timer continues if you leave or close the page.
During the assessment, work independently without this guide, notes, search engines, AI tools or help from another person. Assistive technology is permitted for accessing the assessment. No answer hints or correct-answer explanations are shown. Full-screen focus mode may record interruptions, but cannot disable other apps, extensions or devices.
Your certificate records completion of this course. Assessment results and any review decisions are retained separately in your account. It is not a professional qualification, institutional ethics approval or approval to publish a particular project. Contact info@arforearth.org for questions or accessibility support.
Reviewing a result
If a learning result needs review, contact info@arforearth.org. Authorized administrators may record a grade decision and approve completion. The original assessment result and any review decision are kept in the account record. An active Research Foundations certificate satisfies the project-creation prerequisite; withdrawing it removes that eligibility for new projects.
LESSON 1 OF 8
1. Understand ARforEarth research and choose a useful question
This guide explains the research practice expected of ARforEarth contributors. Read all eight sections before taking the Research Foundations assessment. Passing the course is required before creating a new research project in ARforEarth. It establishes familiarity with these rules; it does not approve a research protocol, guarantee publication, verify a project’s findings or confer professional accreditation.
The requirements below are ARforEarth’s contributor rules for this course. External sources are reading references, not partner organisations or endorsers. Where a project requires specialist supervision, institutional permission or an ethics review, obtain that separately before beginning the relevant work. Ask info@arforearth.org about the current submission route, review capacity or accessibility support.
Learning objectives: explain the difference between a private workspace project and accepted research; define a focused sustainability question; choose an output that fits the question and available evidence.
ARforEarth connects sustainability learning with research and practical action. A useful investigation starts with a question about a defined place, group, service or decision. It can study an environmental, social or economic issue and its connections with the others. Begin by explaining why the question matters to the people affected. A relevant SDG target can help describe the context, but adding an SDG name or logo does not establish impact, UN endorsement or research quality.
Turn an ambition into a workable question. “Make our campus sustainable” is too broad for a small study. “How much avoidable food waste is recorded per lunch served in this canteen during four teaching weeks, and what operational factors might explain the pattern?” identifies an outcome, a boundary and a period. Define what counts as avoidable waste and what a lunch served means before collecting data. Check that the study is feasible with the time, equipment, access and supervision available.
Choose the output to match the purpose. ARforEarth’s published format guidance lists working papers at 3,000–8,000 words, policy briefs at 1,000–2,500 words, research reports at 2,000–6,000 words, case studies at 1,500–5,000 words and impact reports at 1,500–5,000 words. Theses have no fixed word limit in the programme guidance; discuss their scope with the team. Confirm current submission requirements before final formatting. A policy brief should explain a specific decision and its evidence; a case study should make the boundaries of its case clear. Word count does not substitute for a sound method.
Write a short project brief containing the question, context, intended users, proposed output, evidence needed, practical limits and expected risks. Label planned benefits as expectations. Do not write as if results already exist. If the question changes, keep a dated note explaining what changed and why, especially if you have already seen the data.
A workspace project is your private working record. Creating it, uploading a manuscript, completing this course or receiving its certificate does not make the work an ARforEarth publication. Publication requires a separate submission, review and acceptance process. Do not present a private draft as approved, reviewed or published.
Preparation task: write one focused question and one sentence stating what your project will not investigate. Select a format and explain why it suits the intended reader.
Learning objectives: build a traceable source record; distinguish primary evidence from commentary; decide whether two measurements can be compared.
Start by finding what is already known about your question. Prefer original research, the dataset creator’s documentation and official methods for the claim being made. A review can help map a field, while a news story can point to a source; neither removes the need to inspect the original evidence where it is available. Record who produced a source, when it was published or updated, what it studied and how it supports your particular claim.
Maintain a source log with the author or organisation, title, year, stable URL or DOI, access date where useful, relevant page or table, and notes on the method and limitations. Keep enough information for another reader to locate the material. Read beyond a search snippet or AI summary. Check whether the source is a completed study, a preprint, a proposal, an opinion, a correction or a retracted item. Do not describe all of these as equivalent evidence.
Before comparing numbers, check the unit, geography, population, period, denominator, measurement method and treatment of missing data. Two charts can use the same label but measure different things. For example, total electricity used by a building and electricity per occupant answer different questions. A local metric linked to an SDG must not be relabelled as an official UN indicator unless its definition and method genuinely match the relevant indicator metadata.
Cite a source close to the claim it supports. Use one consistent reference style and include a complete reference list. Put copied wording in quotation marks or a clearly labelled quotation, with a source. Paraphrasing requires genuinely expressing the idea in your own words and still crediting its origin. A citation alone does not make unmarked copying acceptable. Credit reused figures, tables, instruments, code and datasets, and check their reuse conditions.
Treat claims generated by AI as unverified leads. A plausible title, DOI or quotation may be wrong. Locate and read the real source before using it; remove a claim or reference you cannot verify. Do not fabricate citations, invent interviews or manufacture results to fill a gap. If evidence is unavailable, say what is missing and how this limits the investigation.
Preparation task: create a five-source evidence table for a possible project. For each source, write the precise claim it can support and one limitation. Keep research notes separate from text copied for quotation.
3. Plan a transparent method and a reproducible record
Learning objectives: connect a question to a suitable method; distinguish an observation from a causal claim; document a data and analysis workflow that another person can understand.
Write the method before collecting or analysing data. State what will be observed, where and when, who or what is included, how observations will be selected, which instruments or sources will be used, and how the results will be summarised. A small case study, a literature review, an interview study and an experiment answer different kinds of questions. Select a method you can explain and carry out competently, and obtain appropriate supervision when the method or setting requires it.
Define variables and units. For a resource study, specify the meter or scale, reading frequency, calibration or checking procedure, the activity denominator and the comparison period. For interviews, document the recruitment approach, consent process, interview topics and analysis method. Explain why the sample is suitable and which people or situations it may miss. A convenience sample of willing respondents must not be described as representative of a whole city without evidence supporting that claim.
Keep raw observations intact. Store analysis in a separate working copy and record every cleaning rule: exclusions, duplicate handling, unit conversions and treatment of missing values. Missing is not zero. Label estimates as estimates and record how they were calculated. If an observation seems unusual, investigate and document it; do not remove it simply because it weakens a preferred result.
A reproducible record includes a dated protocol, a data dictionary, source or instrument details, permitted data, analysis steps or code, relevant software versions, and a short README that explains how to reproduce the reported tables or calculations. Name files consistently and record changes. Someone should be able to understand the workflow even when access to protected raw data must be restricted.
Be careful with causal language. A decline in electricity use after an awareness campaign may also reflect attendance, weather, equipment changes or meter differences. A before-and-after comparison alone rarely isolates the campaign’s effect. Report the observed change and discuss alternative explanations. Do not claim that a correlation proves causation. If a design changes after the study begins, document the change rather than rewriting the original plan as if it had never changed.
Preparation task: outline a one-page method with the sample, measures, units, timing, comparison and analysis. List two alternative explanations for a result you might observe and how your design will examine or acknowledge them.
4. Protect people, communities and sensitive information
Learning objectives: recognise when a project needs additional permission or specialist review; plan informed participation; avoid exposing people through research files or publication.
Before recruitment, observation, recording or data collection, identify possible harm to people, communities, ecosystems and the researchers. Ask what information is actually needed, who may be identifiable, whether a power imbalance could affect participation, and how the work will be supervised. Completing this course is not ethics approval. Obtain required institutional or site permissions and appropriate ethics review before the relevant activity. If you are unsure whether these are needed, pause that activity and consult your supervisor or the ARforEarth team.
For research involving participants, provide a clear explanation of the purpose, activities, expected time, foreseeable risks, intended uses of the information, contact route and any limits on withdrawal. Participation must be voluntary. Do not suggest that access to a service, a grade, employment or a benefit depends on agreeing. Record consent in an appropriate form and respect a refusal. Work involving children, vulnerable groups, health information, high-risk activities or intrusive methods needs suitable expert oversight; the workspace is not a substitute for it.
Collect only the personal information necessary for the stated research purpose. Decide who may access it, how it will be protected, how long it will be retained and how it will be deleted. Keep identifying details separate from analysis data where practical. Replacing a name with a code is pseudonymisation, not a guarantee of anonymity: a small location, rare role or distinctive quotation may still identify someone. Check indirect identifiers before sharing.
Do not upload confidential or identifiable participant material to this general research workspace unless you have established an appropriate, authorised data-handling arrangement for that information. Do not paste protected data into an external AI service. The account workspace can synchronise research records and extracted document text; original uploaded files remain browser-local. Neither arrangement should be treated as an approved repository for sensitive human-participant data. Keep necessary protected originals in the authorised system specified in your project plan.
Open research does not mean publishing every raw file. Share anonymised or aggregated material only where rights, consent and applicable requirements permit it. When data cannot be shared, explain the restriction and provide a safe account of the method or an appropriate access procedure. Respect community context, permission for photographs or quotations, and reasonable expectations of privacy even when material is accessible online.
Preparation task: write a data-handling note covering purpose, minimum information, consent or other required permissions, access, storage, sharing and retention. If the plan cannot protect participants, redesign or stop before collecting data.
5. Work honestly: authorship, plagiarism, conflicts and AI
Learning objectives: apply ARforEarth’s integrity rules; assign human responsibility for a manuscript; disclose support, competing interests and meaningful AI assistance.
ARforEarth requires honest research records and original work with appropriate attribution. Do not invent data, interviews, participants, citations or approvals. Do not alter records or selectively omit evidence to mislead. Clearly identify hypothetical examples, simulations and synthetic data. They can support learning or a declared method, but must never be presented as real field measurements.
Discuss authorship and contributions early. List people who have made a genuine intellectual contribution and can take responsibility for the work; acknowledge other assistance appropriately. Every listed author should review and agree to the submitted version and their contribution statement. Do not add someone for status or remove a qualifying contributor to avoid crediting them. Use accurate affiliations, and do not imply that an institution endorses the work merely because an author is associated with it.
Disclose funding, material support and interests that could reasonably influence the work or a reader’s interpretation. This can include employment, relationships with a studied organisation, a financial interest in an intervention, or advocacy connected to the question. A disclosed interest is information for assessment; hiding it prevents an informed review. Report the funder’s or partner’s role in design, access, analysis and publication decisions.
AI assistance does not transfer responsibility away from the human authors. If AI materially assists drafting, translation, coding, analysis or image creation, include a disclosure stating the tool, its purpose, the parts affected and how a human checked the output. Preserve enough detail to explain a method that depends on AI. Verify factual statements, calculations, code behaviour and every reference against real evidence. An AI system is not an author and cannot confirm participant consent or ethical approval.
Do not submit copied or AI-produced material as if it were your unassisted original research. Do not upload confidential inputs to a tool without appropriate permission and safeguards. A grammar or accessibility aid may support communication, but the author remains accountable for meaning and accuracy. During the Research Foundations assessment, independent work is required: do not consult the guide, another person, answer-sharing services or AI tools. Assistive technology used to access the interface is permitted; ask the team if an adjustment is needed.
Preparation task: draft an authorship and contribution statement, an interests statement and an AI-use statement for a hypothetical project. If there is no relevant support, interest or AI use, state that accurately rather than leaving the reader to guess.
Learning objectives: organise a clear research manuscript; distinguish results from interpretation; report uncertainty and boundaries without overstating impact.
A reader needs to understand the question, the evidence and the reasoning connecting them. Use a descriptive title and a concise abstract covering the purpose, method, main result and limits. The introduction should explain the problem and existing evidence. The methods should make the investigation understandable. Present results before discussing what they may mean, and finish with conclusions that stay within the study’s scope. Adapt this structure to the selected format rather than forcing a policy brief to read like a laboratory report.
Include authors and contributions, affiliations where relevant, references, support and competing interests, ethics and consent information where applicable, a data/code availability statement, and a disclosure of meaningful AI use. Add appendices for necessary detail. Do not expose participant identities in acknowledgements, screenshots, file names, appendices or document properties. Confirm permission and licensing for every third-party figure, photograph or substantial reused item.
Separate observations, calculations, estimates and interpretation. State units and denominators in tables and charts. For example, 50 kg falling to 40 kg is a 10 kg decrease, or 20% of the original 50 kg. That arithmetic does not by itself show why the change happened. State the comparison periods and whether activity, measurement conditions or disposal routes changed. Do not turn an estimated benefit into a measured result.
Describe missing data, sample limitations, uncertainty, possible bias and alternative explanations. A small voluntary sample can reveal useful experiences while failing to represent everyone. A finding with no clear improvement is still a finding. Report inconvenient or null results and explain deviations from the planned method. Do not claim a project “proves” a universal solution on the strength of one pilot.
For sustainability claims, specify the boundary. An operational electricity estimate does not cover manufacturing or disposal unless those are included. A carbon estimate needs a documented emissions factor, geography, period and calculation method. Count project activities, outputs and outcomes separately: a workshop delivered is not automatically a reduction in emissions or an improvement in well-being. A DOI, certificate, SDG connection or attractive chart cannot strengthen an unsupported claim.
Preparation task: create a submission checklist for your format. Write a three-sentence result statement: what was observed, how it was measured, and what remains uncertain. Have another reader check whether every conclusion can be traced to evidence.
7. Submit, respond to review and publish responsibly
Learning objectives: describe the ARforEarth submission and review stages; respond to criticism with evidence; explain what acceptance, a repository record and a DOI do and do not establish.
When the manuscript and supporting record are ready, contact info@arforearth.org or use the research inquiry route to confirm the current submission instructions and review capacity. The published process involves initial screening, community feedback, revision and a publication decision. A private workspace upload is not a submission, and submission is not acceptance. Do not promise a review deadline, publication date, DOI or indexing outcome before the team has confirmed it.
Screening can consider fit, completeness, integrity, permissions, clarity and whether claims are supportable. Community review is constructive feedback and is different from journal peer review. Describe the actual process accurately in the manuscript and any promotion. Do not advertise work as journal peer-reviewed, institutionally endorsed or professionally accredited merely because ARforEarth has reviewed or hosted it.
Keep a response log for revision. For each comment, state what you changed and where, or explain with evidence why you disagree. Respectful disagreement is allowed; silently ignoring a substantive concern is not a response. Preserve dated versions so the team can see the revision history. Authors remain responsible for their methods and claims after reviewers have commented.
Accepted work may be shared through Zenodo with a DOI, subject to the team’s decision and the repository’s requirements. Confirm author names, title, version, description, funding details, related records, permissions and the licence before publication. Publish only material you have the right to share. An open licence permits specific reuse under its terms; it is not a way to erase someone else’s rights or participant confidentiality.
A DOI provides a persistent identifier for a research object. It is not a quality score, proof that findings are true, evidence of UN endorsement or a guarantee of indexing. Distinguish a particular version of a work from a record representing its versions. Cite the version actually used when that distinction matters. External platforms decide their own availability, indexing and display.
Preparation task: draft a cover note stating the research question, selected format, originality, author approval, disclosures and any restricted data. Then respond to one imagined reviewer concern with a specific revision and a reasoned explanation.
8. Maintain your research record and earn your course certificate
Learning objectives: handle an error after submission or publication; distinguish research eligibility from publication approval; prepare for an independent assessment of these rules.
Research responsibility continues after a file is submitted. Keep the permitted evidence, analysis and revision record for the period agreed in your project plan, subject to consent and applicable requirements. Keep separate backups of original files. The ARforEarth account workspace synchronises records and extracted document text, while original uploads remain in the browser where they were added. A successful account save does not prove that the original file is stored in the cloud.
If you discover an error, tell the ARforEarth team promptly. Explain the affected version, what is wrong, how you discovered it, the likely effect on results and your proposed correction. Preserve the earlier record and make corrections traceable. Do not quietly replace a published file to conceal a material change. A serious integrity or privacy problem may require pausing dissemination and discussing correction, withdrawal or other action with the team and repository.
Use accurate claims when sharing the work. Distinguish a draft, submitted manuscript, accepted work and published version. Give the relevant link and version where available. Explain the method and limitations to a non-specialist audience. Correct a misleading public claim as well as the underlying file. Respect requests and restrictions that protect participants or confidential material.
Before creating a new research project in ARforEarth, sign in, enrol in ARforEarth Research Foundations, read and complete all eight required lessons, and pass its assessment. A server-issued course certificate records that the account met those learning requirements. Private practice-course completion records do not satisfy this prerequisite. The research certificate permits the project-creation step; it does not approve human-participant work, grant permission to use someone else’s data or guarantee publication.
Assessment rules: prepare using this guide, then work independently without the guide, notes, AI tools, another person or answer-sharing services during the attempt. The assessment selects 12 questions, allows 30 minutes and requires at least 80% correct, which means at least 10 correct answers out of 12. Up to three new attempts may be started in a rolling 24-hour period. Review the guide between attempts. The assessment supplies no hints or answer key; the result reports your overall outcome.
The website may request full-screen assessment mode and record when its assessment page loses focus. These are limited integrity measures, not a secure exam browser. A website cannot block other apps, AI services or a second device, and a focus change alone does not prove cheating. Use a stable connection and minimise interruptions. Assistive technology is permitted for accessing the assessment. Contact info@arforearth.org for accessibility help or a technical problem; do not share your password or private authentication details.
Final preparation: read the entire guide, check that you can explain the process in your own words, and resolve any accessibility or technical concerns before starting. Your certificate is an ARforEarth course-completion record; it is not a degree, professional licence, UN qualification or independent verification of your research.
