Skip to main content
v2026.11,986 entries · CC-BY 4.0

Value Sensitive Design: Research Software and Systems Guide

A methodological guide to Value Sensitive Design for research software, covering IEEE 7000 integration, tripartite investigations, and algorithmic audits.

Written and maintained by CASRAI Editorial Board

Published · Last updated

On this page

Value Sensitive Design (VSD) is a theoretically grounded software and systems engineering methodology that accounts for human values in a principled, comprehensive manner throughout the entire technology lifecycle. Originated by Batya Friedman, Peter H. Kahn Jr., and Alan Borning at the University of Washington, VSD unites philosophical conceptual inquiry, empirical stakeholder research, and technical design investigations into an iterative tripartite process. In modern computer science and research computing, VSD enables research teams to translate ethical concerns—such as privacy, algorithmic fairness, human autonomy, and accountability—into verifiable software requirements, data governance pipelines, and systems architecture.

Within modern computer science, researchers routinely develop algorithms, autonomous models, and data-intensive pipelines that exert direct influence on individuals and society. Traditional software engineering methodologies prioritize functional performance, operational efficiency, and usability heuristics, often treating ethical questions as downstream regulatory afterthoughts. Value Sensitive Design fundamentally alters this dynamic by establishing human values as first-class design constraints from the earliest conceptualization phases. Rather than treating technology as morally neutral, VSD adopts an interactional stance: technologies actively structure human possibilities, while human values simultaneously shape how technologies are adopted, subverted, or governed.

What it is used for in research

In academic and industrial research, Value Sensitive Design provides an operational framework for identifying, anticipating, and resolving ethical tensions in computational systems. Researchers apply VSD across machine learning pipelines, health informatics, automated decision systems (ADS), digital surveillance platforms, and collaborative research data infrastructure. The methodology prevents common failure modes where software unintentionally marginalizes vulnerable groups, compromises participant autonomy, or introduces systemic bias.

The core engine of VSD is its tripartite methodology, which mandates three continuous, iterative modes of investigation:

  • Conceptual Investigations: Philosophically grounded inquiries that define the specific human values implicated by the system (e.g., human welfare, privacy, freedom from bias, universal usability, trust, informed consent, and environmental sustainability). Conceptual analysis maps the boundary between direct stakeholders (those who directly interact with the software interface or administrative tooling) and indirect stakeholders (individuals, organizations, or ecosystems impacted by the system’s decisions or data outputs, even if they never touch the interface).
  • Empirical Investigations: Social scientific inquiries—including contextual interviews, surveys, ethnographies, and stakeholder co-design sessions—that examine how real human participants perceive, prioritize, and experience specific values in operational contexts. Empirical work reveals cultural variations, power imbalances, and value conflicts that theoretical analysis alone cannot anticipate.
  • Technical Investigations: Software engineering and algorithmic analyses focused on the system itself. Technical investigations identify how existing architectural mechanisms (e.g., database schemas, authentication layers, loss functions, and classification thresholds) support or inhibit human values, and proactively design new technical mechanisms that enforce targeted values.

A critical contribution of VSD to research computing is its formalization under international standards. In 2021, the IEEE Standards Association published IEEE 7000-2021 (co-released internationally as ISO/IEC/IEEE 24748-7000:2022), titled Standard Model Process for Addressing Ethical Concerns During System Design. IEEE 7000 standardizes VSD principles into formal systems engineering lifecycles. It establishes rigorous processes for Concept Exploration, the derivation of Ethical Value Requirements (EVRs), and continuous values traceability linking philosophical ethical principles directly to software architecture, unit tests, and post-deployment monitoring.

Researchers in computational domains utilize VSD to address value tensions that cannot be resolved by mathematical optimization alone. For example, in predictive analytics and algorithmic modeling, research teams must balance statistical predictive accuracy against algorithmic fairness and individual privacy. By applying VSD alongside quantitative tools like AI Fairness 360 (AIF360), teams move beyond simplistic metrics to examine who bears the risks of model false positives and how marginalized communities are protected. Furthermore, examining ethical tensions through VSD helps distinguish between institutional compliance mandates and broader ethical obligations, a critical distinction explored in CASRAI’s analysis of research ethics.

How to install or access it

Implementing Value Sensitive Design in research software involves accessing official methodological toolkits from the originating academic laboratory and installing open-source software libraries designed for technical ethical auditing and fairness evaluations.

Official Methodological Toolkits

The University of Washington Value Sensitive Design Lab (VSD Lab) provides official, verified toolkits for conducting conceptual and empirical stakeholder investigations:

  • The Envisioning Cards Toolkit (2nd Edition, 2024): Designed by Batya Friedman, David G. Hendry, and colleagues at the University of Washington, this official toolkit comprises 42 cards licensed under Creative Commons (CC BY-NC-ND 4.0). The cards are organized across five thematic suits: Stakeholders, Time, Values, Pervasiveness, and Multi-lifespan design. Official PDF editions are accessible directly from the University of Washington VSD Lab repository at https://envisioningcards.vsdesign.org and https://vsdesign.org/toolkits/envisioningcards/.
  • Value Dams and Flows Matrix: The official stakeholder prioritization heuristic developed by Miller et al. and the VSD Lab, accessible through the VSD Lab publications repository at https://vsdesign.org/publications/.
  • IEEE 7000-2021 Process Standard: Official engineering standard documentation is maintained and accessible through the IEEE Standards Association at https://standards.ieee.org/ieee/7000/7057/.

Software Environment Setup

For computational researchers embedding VSD into Python-based data pipelines and software repositories, two key open-source packages operationalize technical investigations: deon (a command-line ethics audit checklist generator) and fairlearn (an algorithmic fairness evaluation and mitigation toolkit).

Install both packages using the official Python Package Index (PyPI) or Conda package managers:

# Create and activate an isolated research environment
python -m venv vsd-env
source vsd-env/bin/activate

# Install deon CLI checklist generator
pip install deon

# Install fairlearn and core dependencies
pip install fairlearn pandas scikit-learn

Verify that both tools are installed correctly and display expected version strings:

# Check deon CLI execution
deon --version

# Verify Fairlearn Python package import
python -c "import fairlearn; print('Fairlearn version:', fairlearn.__version__)"

Step-by-step task

The following numbered walkthrough demonstrates how to execute an end-to-end Value Sensitive Design investigation on a machine learning predictive software pipeline. This task integrates conceptual stakeholder mapping, automated repository ethics auditing, and quantitative fairness evaluations.

  1. Map Direct and Indirect Stakeholders:
    Before writing algorithmic code, convene an initial conceptual investigation session. Construct a stakeholder classification ledger distinguishing:

    • Direct Stakeholders: End-users interacting with the software dashboard, research administrators configuring parameters, and system engineers deploying code.
    • Indirect Stakeholders: Human subjects evaluated by the algorithm, demographic subgroups whose historical data trains the system, and public communities affected by downstream automated resource allocation.

    Record all stakeholder groups in a project charter document within the software repository.

  2. Formulate Ethical Value Requirements (EVRs):
    Translate high-level philosophical values into verifiable software engineering requirements adhering to IEEE 7000. For instance, translate the value “Freedom from Algorithmic Bias” into the technical requirement: “The false positive rate (FPR) disparity across protected demographic classes shall not exceed 0.05 at the 0.5 decision threshold.” When dealing with user data access, ensure clear boundaries between passive terms and explicit permissions, examining mechanisms of informed consent in all data harvesting workflows.
  3. Initialize the Project Ethics Audit Checklist via Deon:
    Generate an operational ethics audit file directly in your project root using the Deon CLI. Deon automatically structures the audit across five lifecycle stages: Data Collection, Storage, Analysis, Modeling, and Deployment:

    # Generate a Markdown checklist in the project repository
    deon -f markdown -o ETHICS.md

    Open ETHICS.md and systematically document the justification for each checkpoint. Commit this file to version control alongside code commits to maintain a tamper-evident audit history.

  4. Quantify Algorithmic Disparities Using Fairlearn:
    In your evaluation script, load test set predictions, ground-truth outcomes, and the sensitive demographic feature column. Utilize Fairlearn’s MetricFrame to evaluate whether the software satisfies your EVRs:

    import pandas as pd
    from sklearn.metrics import accuracy_score, precision_score
    from fairlearn.metrics import MetricFrame, selection_rate, false_positive_rate
    
    # Evaluate predictive metrics across demographic groups
    metric_frame = MetricFrame(
        metrics={
            "accuracy": accuracy_score,
            "selection_rate": selection_rate,
            "false_positive_rate": false_positive_rate
        },
        y_true=y_test,
        y_pred=y_pred,
        sensitive_features=sensitive_feature_test
    )
    
    # Inspect disaggregated group performance and maximum disparity
    print(metric_frame.by_group)
    print(metric_frame.difference(method="between_groups"))
  5. Execute Algorithmic Mitigation and Value Re-balancing:
    If the technical investigation reveals that false positive disparities exceed the acceptable EVR threshold, apply an in-processing mitigation algorithm (such as Fairlearn’s ExponentiatedGradient with EqualizedOdds) or post-processing threshold adjustments. Re-run the MetricFrame to confirm that disparity metrics now fall within acceptable bounds without catastrophically degrading overall utility.
  6. Apply the Value Dams and Flows Heuristic:
    In a collaborative workshop with stakeholder representatives, evaluate candidate software features using the Value Dams and Flows heuristic:

    • Value Dam: Any proposed feature or algorithmic behavior strongly opposed by even a small minority of stakeholders (e.g., over 10% strong opposition) constitutes a Dam and is categorically removed from the software architecture.
    • Value Flow: Features that receive widespread positive support (over 50% stakeholder endorsement) and violate no Value Dams are designated as Flows and prioritized for development sprints.
  7. Archive Values Traceability in Repository:
    Export the finalized EVR mapping, Deon checklist, and Fairlearn disparity reports into an /ethics directory in the software repository. If publishing the software in scholarly venues such as the Journal of Open Source Software (JOSS), reference these files within your paper’s statement of need and software documentation.

Output and how to read it

Executing Value Sensitive Design produces three primary operational outputs: the Deon Ethics Audit Report, the Fairlearn Disparity Metric Table, and the Values Traceability Matrix. Understanding how to interpret each output ensures that research teams make informed, defensible engineering decisions.

1. Deon Ethics Audit Report (ETHICS.md)

The generated Deon Markdown report structures the project’s ethical posture into distinct sections:

  • Data Collection: Documents whether informed consent was obtained, whether secondary data uses were disclosed, and whether data collection biases were evaluated.
  • Data Storage: Confirms data minimization principles, encryption standards at rest and in transit, and role-based access control policies.
  • Modeling and Evaluation: Verifies that baseline comparisons were established, performance across demographic slices was measured, and proxy variables for protected attributes were eliminated.
  • Deployment: Audits rollback procedures, human-in-the-loop oversight mechanisms, and post-deployment drift detection schedules.

2. Fairlearn Disparity Metric Output

Fairlearn’s MetricFrame produces a disaggregated evaluation matrix. An illustrative output table for an automated classification system illustrates how these metrics appear:

Subgroup Category Sample Count (n) Accuracy Selection Rate False Positive Rate (FPR)
Overall Population 10,000 0.882 0.320 0.084
Group Alpha (Historical Majority) 6,500 0.895 0.354 0.062
Group Beta (Protected Minority) 2,500 0.852 0.248 0.141
Group Gamma (Protected Minority) 1,000 0.870 0.280 0.095
Maximum Difference (Between Groups) — 0.043 0.106 0.079

When reading this output, researchers must examine three critical disparity metrics:

  • Selection Rate Disparity: The difference between the highest and lowest group selection rates (0.354 – 0.248 = 0.106). In regulatory and research contexts, the four-fifths rule (selection rate ratio ≥ 0.80) serves as a standard heuristic. Here, the ratio is 0.248 / 0.354 = 0.700, indicating an adverse impact that violates the selection rate criterion.
  • False Positive Rate Disparity: Group Beta experiences a false positive rate of 14.1% compared to 6.2% for Group Alpha, yielding a disparity of 0.079 (7.9 percentage points). In punitive or high-stakes screening software, this disparity directly violates EVRs mandating equalized odds.
  • Accuracy Spread: An accuracy difference of 0.043 indicates that model reliability varies across cohorts, necessitating data augmentation or re-sampling in Group Beta before deployment.

3. Values Traceability Matrix

The Values Traceability Matrix links each identified human value through its conceptual definition, corresponding EVR, architectural implementation, and verification test. For example, the value “Privacy” traces to the EVR “Zero exposure of raw IP addresses in application logs”, which traces to an architectural hashing middleware, which traces to an automated unit test in the CI/CD pipeline verifying that log outputs match cryptographic hashes.

The data table below compares the structural attributes and life-cycle phases of Value Sensitive Design against five alternative responsible technology frameworks.

Troubleshooting

Software and systems researchers frequently encounter conceptual and technical hurdles when operationalizing Value Sensitive Design. Below are five common issues and their verified remedies.

1. Indirect Stakeholder Blind Spots

Problem: Engineering teams focus exclusively on direct end-users (e.g., lab technicians entering data), completely overlooking indirect stakeholders (e.g., research participants whose records are analyzed). Consequently, the software introduces severe harms to unrepresented groups.

Remedy: Deploy the Stakeholders suit of the VSD Envisioning Cards toolkit during sprint planning. Explicitly prompt the team with cards such as Indirect Stakeholders, Excluded Stakeholders, and Children or Vulnerable Populations. Require team members to document at least three non-user populations impacted by the system’s inputs, processing, or outputs before approving the software architecture.

2. Vague or Conflicting Ethical Value Definitions

Problem: Teams define values using high-level slogans without operational criteria, leading to design stalemates when engineers cannot determine what code to write.

Remedy: Apply the IEEE 7000 value conceptualization protocol. Deconstruct each value into specific operational indicators, observable behaviors, and measurable thresholds. For Fairness, specify whether the system optimizes for demographic parity, predictive parity, or equalized odds. Recognize that mathematical incompatibility theorems prevent satisfying all fairness definitions simultaneously, requiring an explicit, documented trade-off decision.

3. The Static Checklist Compliance Trap

Problem: Teams treat the Deon ETHICS.md checklist as a one-time bureaucratic requirement completed at project inception, allowing code changes to violate ethical requirements unnoticed during later development.

Remedy: Embed ethical verification into the automated continuous integration (CI) pipeline. Use pre-commit hooks that require updating audit notes whenever core data loading or model training files are modified. Implement automated regression tests using Fairlearn that fail the build if disparity metrics exceed pre-defined EVR thresholds.

4. Value Tension Deadlocks

Problem: Development halts because two fundamental values directly conflict—for instance, maximizing data privacy through aggressive differential privacy noise versus maximizing scientific reproducibility and analytical utility.

Remedy: Apply the Value Dams and Flows method. First, establish whether either side of the tension constitutes a Dam (an absolute breach of participant trust or legal mandate). If neither is an absolute dam, construct an explicit Value Hierarchy per IEEE 7000 Clause 6. Rank the values transparently based on stakeholder input, select a compromise threshold, and publicly document the rationale for the selected equilibrium.

5. Small Subgroup Sample Sizes in Fairlearn MetricFrame

Problem: When evaluating intersections of sensitive features, certain subgroup buckets contain very few observations (sample size below 30), producing NaN disparity scores or extreme variance that distorts audit metrics.

Remedy: In your evaluation code, set the MetricFrame parameter sample_params or filter out underpowered strata while flagging them in your audit report. Calculate Wilson score confidence intervals or bootstrap standard errors around group-level metrics rather than relying solely on point estimates. Never treat small subgroup counts as evidence of fairness; annotate data sparsity as an area requiring targeted recruitment or empirical follow-up.

Free alternatives

While Value Sensitive Design provides an exceptionally comprehensive framework for integrating human values across the entire lifecycle of software and systems, research teams frequently evaluate alternative responsible design methodologies. The table below compares VSD with five prominent alternatives across structural dimensions.

Framework Primary Focus Stakeholder Participation Mode Core Mechanism Free Toolkits and Standards
Value Sensitive Design (VSD) Principled accounting of human values across technology lifecycles Empirical investigations with direct, indirect, and excluded stakeholders Iterative Tripartite Methodology (Conceptual, Empirical, Technical) Envisioning Cards (CC BY-NC-ND 4.0), Deon CLI, Fairlearn
IEEE 7000-2021 (VBE) Standardized systems engineering model for addressing ethical concerns Structured stakeholder engagement and expert ethical risk assessments Ethical Value Requirements (EVRs) and end-to-end values traceability IEEE Standards Association open summaries, ISO/IEC/IEEE 24748-7000
Participatory Design (PD) Democratizing workplace technology and empowering frontline workers Deep co-design where end-users act as equal design partners Mutual learning, design games, paper prototyping, and iterative co-creation ACM PDC community toolkits, Participatory Design toolcards (Open Access)
Critical Technical Practice (CTP) Reflexive critique and reimagining of core technical assumptions Developer self-reflection and critical scholarly engagement Double-movement: building working systems while dismantling their ideological premises Critical code studies frameworks, open-source subversive computing tools
Responsible Research and Innovation (RRI) Aligning research and innovation outcomes with societal needs and values Multi-actor public engagement (citizens, policymakers, industry, researchers) AREA Framework (Anticipate, Reflect, Engage, Act) across 4 dimensions European Commission RRI Toolkit, Societal Readiness Level (SRL) assessments
Design Justice Dismantling structural inequalities and centering marginalized communities Community-led co-design prioritizing those most harmed by systems 10 Design Justice Principles challenging the ‘matrix of domination’ Design Justice Network Principles, community-authored zines and audit guides

Choosing among these frameworks depends on the institutional setting, regulatory context, and research goals:

  • Choose VSD when designing computational architectures, software platforms, or data pipelines that involve complex stakeholder networks and require simultaneous philosophical rigor, empirical user research, and technical software engineering.
  • Choose IEEE 7000 when operating within formal systems engineering environments, defense projects, medical device manufacturing, or aerospace software where auditable standards compliance and rigorous requirements traceability are contractually required.
  • Choose Participatory Design when the primary objective is workplace empowerment or clinical workflow redesign, and where frontline workers can actively participate in iterative design sprints.
  • Choose Design Justice when building civic technology, community organizing software, or algorithmic tools that directly address systemic oppression, ensuring that power over design decisions remains with marginalized communities.

Frequently asked questions

What is the tripartite methodology in Value Sensitive Design?

The tripartite methodology is the operational core of Value Sensitive Design, integrating conceptual, empirical, and technical investigations into an iterative, mutually informing loop. Conceptual investigations philosophically define the values at stake and identify both direct and indirect stakeholders. Empirical investigations deploy qualitative and quantitative social science methods to evaluate how stakeholders prioritize and experience these values in real-world contexts. Technical investigations analyze how hardware and software properties support or hinder human values and proactively engineer technical mechanisms to uphold them.

How does Value Sensitive Design differ from traditional user-centered design?

Traditional User-Centered Design (UCD) focuses primarily on direct users, optimizing for immediate usability, task completion time, efficiency, and user satisfaction. In contrast, Value Sensitive Design expands the scope of analysis in three fundamental ways: it accounts for indirect stakeholders who are impacted by the technology without operating it; it prioritizes enduring ethical and moral values (such as justice, privacy, and human dignity) over mere functional convenience; and it addresses long-term systemic impacts across multiple generations rather than immediate interaction cycles.

What are value dams and value flows in VSD?

Value Dams and Value Flows constitute an empirical heuristic developed within VSD to resolve stakeholder value tensions and prioritize design features. A “Value Dam” represents any proposed technical feature, policy, or algorithmic behavior that elicits strong opposition from even a small proportion of stakeholders (traditionally 10% or more); such features are treated as non-negotiable boundaries and removed. Conversely, a “Value Flow” is a feature that receives widespread positive endorsement from a substantial majority (50% or more) without triggering any Value Dams, designating it as a high-priority feature for implementation.

Who qualifies as an indirect stakeholder in software research?

In Value Sensitive Design, an indirect stakeholder is any individual, demographic group, institution, or ecological community that is affected by the operation, outputs, or societal presence of a system, despite never directly interacting with its physical hardware or user interface. Examples in research computing include patients whose health outcomes are influenced by an automated diagnostic tool, citizens whose creditworthiness is scored by a machine learning model, and downstream communities impacted by the energy consumption of large-scale computational infrastructure.

How does IEEE 7000 operationalize Value Sensitive Design in engineering projects?

IEEE 7000-2021 formalizes VSD principles into standard systems engineering workflows by establishing a systematic five-stage process: Concept Exploration, Context of Use analysis, Ethical Value Elicitation, Ethical Value Requirements (EVRs) formulation, and Values Traceability management. It requires engineering organizations to translate abstract ethical principles into auditable, testable software requirements, ensuring that values such as transparency, accountability, and fairness are tracked and verified through the entire software development lifecycle.

Can Value Sensitive Design be integrated into automated CI/CD pipelines?

Yes. In modern computational research, technical investigations within VSD are operationalized inside continuous integration and continuous deployment (CI/CD) pipelines. Research teams integrate automated ethics checklist verification tools like Deon via pre-commit hooks and embed programmatic disparity regression suites using libraries like Fairlearn. If a model retrained on fresh data produces a demographic parity difference or false positive disparity exceeding the thresholds specified in the project’s Ethical Value Requirements, the automated CI/CD build fails, preventing the deployment of biased code.

Sources

  • Friedman, B., Kahn, P. H., and Borning, A. (2006). Value Sensitive Design and Information Systems. In P. Zhang and D. Galletta (Eds.), Human-Computer Interaction and Management Information Systems: Foundations (pp. 348–372). M.E. Sharpe. https://vsdesign.org/publications/ (Accessed: 2026-11-16)
  • Value Sensitive Design Research Lab. (2024). Value Sensitive Design Theory, Methods, and Projects. University of Washington Information School. https://vsdesign.org (Accessed: 2026-11-16)
  • Friedman, B., Nathan, L., Kane, S., Lin, J., Yoo, D., Logler, N., Ballard, S., and Hendry, D. G. (2024). Envisioning Cards: A Value Sensitive Design Toolkit (2nd Edition). University of Washington. https://envisioningcards.vsdesign.org (Accessed: 2026-11-16)
  • IEEE Standards Association. (2021). IEEE 7000-2021: IEEE Standard Model Process for Addressing Ethical Concerns During System Design. IEEE Standards Association. https://standards.ieee.org/ieee/7000/7057/ (Accessed: 2026-11-16)
  • DrivenData. (2024). Deon: An Ethics Checklist for Data Scientists. DrivenData Open Source. https://deon.drivendata.org (Accessed: 2026-11-16)
  • Fairlearn Contributor Community. (2024). Fairlearn: A Python Toolkit for Assessing and Mitigating Algorithmic Unfairness. Fairlearn Project. https://fairlearn.org (Accessed: 2026-11-16)

Data: Comparison of Responsible Engineering and Ethical System Design Frameworks

Comparative analysis of six prominent ethical and responsible technology design frameworks for computer science and software research, evaluated across origin, core methodology, stakeholder model, and implementation tooling.

Period covered: 2021/2026. Coverage: Global.

Comparison of Responsible Engineering and Ethical System Design Frameworks
FrameworkStandardizing Body or OriginPrimary Lifecycle PhaseCore Methodological MechanismStakeholder ModelSoftware and CLI Tooling
Value Sensitive Design (VSD)University of Washington (Friedman et al.)Full Lifecycle (Concept to Deployment)Tripartite Iteration (Conceptual, Empirical, Technical)Direct and Indirect StakeholdersEnvisioning Cards, Deon CLI, Fairlearn
IEEE 7000-2021 (Value-Based Engineering)IEEE Standards Association / ISO/IECConcept Exploration and Requirements EngineeringEthical Value Requirements and Values TraceabilityDirect, Indirect, and Systemic StakeholdersIEEE 7000 Process Audits, ISO/IEC 24748-7000
Participatory Design (PD)Scandinavian School / ACM PDCRequirements and PrototypingDemocratic Co-design and Mutual LearningEnd-Users and Frontline WorkersCo-design Workshops, Paper Prototyping Toolkits
Critical Technical Practice (CTP)MIT AI Lab (Agre) and Critical DesignAlgorithmic Architecture and System InceptionReflexive Critique of Technical AssumptionsSystem Developers and Cultural SubjectsCritical Code Studies, Reflective Code Reviews
Responsible Research and Innovation (RRI)European Commission / Horizon FrameworksResearch Governance and CommercializationAREA Framework (Anticipate, Reflect, Engage, Act)Institutional, Societal, and Policy StakeholdersRRI Toolkits, Societal Readiness Level (SRL) Audits
Design Justice FrameworkDesign Justice Network (Costanza-Chock)Design Ideation and Structural GovernanceCentering Marginalized Communities in DesignSystemically Excluded and Historically Marginalized GroupsDesign Justice Principles Audits, Community Toolkits

Sources

Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.

Follow CASRAI

Research-administration guidance, standards updates and independent tool reviews.

Ask CASRAI · free to try

Ask about Value Sensitive Design: Research Software and Systems Guide

2 free questions a day, no account.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →