Can Generative AI Evaluate Whether a Proposed Research Design Is Feasible?

Can AI Tell Whether Your Research Design Can Actually Be Implemented?

You have developed a research proposal with an ambitious methodology. It includes several institutions, repeated measurements, comparison groups, and sophisticated analytical procedures. Generative AI describes the design as rigorous and recommends proceeding.

But you have only one semester to collect data, limited funding, and no confirmed access to several participating institutions.

Can AI recognize that a design may be methodologically defensible yet practically impossible to implement?

Research feasibility concerns whether a proposed investigation can be carried out under actual conditions. A study may have a convincing theoretical rationale and an appropriate design while depending on recruitment targets, resources, permissions, or procedures that cannot realistically be achieved.

These constraints matter because an unworkable design may lead to incomplete data, protocol deviations, inadequate comparisons, or conclusions that are weaker than the original proposal anticipated.

AI Can Help Assess Feasibility, but It Cannot Verify Conditions It Does Not Know

How AI Can Help Evaluate the Practical Requirements of a Research Design

What Does Research Design Feasibility Mean?

Feasibility concerns whether the proposed research can be conducted as intended within the available conditions and constraints.

These conditions may include the availability of eligible participants, institutional permissions, financial resources, qualified personnel, measurement instruments, technology, and sufficient time for implementation and analysis.

Eldridge and colleagues (2016) describe feasibility research as investigating whether a future study can be undertaken and how it should be conducted.

Although their framework was developed in the context of preparation for randomized trials, the broader principle is useful across research settings: uncertainties about implementation should be examined before a definitive investigation begins.

AI may help identify these uncertainties, but the researcher must determine whether they actually exist and how they affect the study.

How Is Feasibility Different From Methodological Appropriateness?

Consider a randomized trial investigating whether AI tutoring improves mathematics achievement.

Random assignment may be methodologically appropriate for estimating an intervention effect. However, the trial may be infeasible if the researcher cannot assign students to instructional conditions or obtain approval to implement the intervention.

Conversely, a cross-sectional survey may be easy to administer but incapable of establishing the causal effect the research question seeks.

Feasibility and methodological fit should therefore be evaluated separately before selecting a final design.

What Feasibility Dimensions Can AI Help Examine?

These dimensions are interdependent. Recruitment delays may affect the timeline, while limited funding may constrain the number of sites or measurements that can be included.

Can AI Determine Whether the Proposed Sample Size Is Feasible?

AI can compare a proposed sample size with the accessible population, recruitment period, anticipated participation rate, and available resources.

However, it should not assume that every eligible person will participate or that a recruitment target can be met simply because the institution has sufficient enrollment.

Suppose a researcher needs 200 participants but has access to only 120 potentially eligible students. The proposed recruitment target is not achievable within that population without changing the sampling frame or other conditions.

More subtle problems arise when the eligible population is large but participation rates are uncertain.

AI may help construct recruitment scenarios, but the assumptions should be informed by previous studies, institutional records, or pilot recruitment evidence where available.

The required sample size itself must also be justified through appropriate methodological considerations. Feasibility does not justify arbitrarily reducing a sample below what the research question and intended precision require.

Can AI Help Evaluate Recruitment Timelines?

Yes. AI may help break recruitment into operational stages and identify dependencies that could delay data collection.

For example, recruitment may depend on ethics approval, institutional authorization, classroom access, informed consent procedures, and participant availability.

A proposed timeline that assumes all these activities can occur simultaneously may be unrealistic.

Researchers should distinguish tasks that can proceed in parallel from those that require prior approval or completion of another activity.

AI-generated timelines should be treated as planning estimates until the relevant institutions and personnel confirm their feasibility.

Can AI Identify Hidden Costs in a Research Design?

AI can help researchers identify expense categories that may be overlooked in an initial budget.

For an educational technology intervention, these may include software subscriptions, devices, internet access, technical support, assessment materials, data storage, participant communication, transcription, and research assistance.

However, AI should not invent current prices or assume that software licenses permit research use.

Actual costs and contractual conditions must be verified through appropriate providers and institutional arrangements.

Researchers should also consider whether the design requires expertise that is not currently available within the research team.

Why Is Feasibility Particularly Important in Educational Technology Research?

Educational technology studies often depend on infrastructure and institutional practices that researchers do not fully control.

For example, a study evaluating a commercial generative AI system may depend on platform availability, account restrictions, licensing conditions, institutional data protection rules, and changes to the system during data collection.

Students may also differ in device access, connectivity, and familiarity with the technology.

These conditions can influence both whether the study can proceed and how consistently the intervention is implemented.

AI may identify potential risks, but it cannot confirm platform stability, local connectivity, or institutional approval without verified information.

Researchers should examine whether field-specific constraints have been overlooked during design selection.

Can AI Assess Ethical Feasibility?

AI can identify potential ethical questions involving informed consent, confidentiality, participant burden, intervention access, and data protection.

For example, a study requiring students to submit identifiable academic work to an external AI platform may raise questions about data governance and institutional permissions.

Similarly, withholding required educational support to create a control group may be ethically problematic.

However, AI cannot grant ethics approval or determine compliance with every applicable institutional or legal requirement.

Researchers must consult responsible ethics bodies and relevant institutional policies before implementation.

Can a Design Be Feasible but Still Scientifically Weak?

Yes. Feasibility does not establish that a design can answer the research question.

A researcher may find it easy to recruit students for a cross-sectional survey about AI use and academic performance. The survey may be completed on time and within budget.

However, if the research question asks whether AI use causes improvements in performance, the design may not support the intended causal conclusion without additional identification assumptions and evidence.

The study is operationally feasible but potentially misaligned with its inferential objective.

Researchers should therefore examine whether the design supports the proposed conclusions rather than selecting the easiest available procedure.

Can AI Recommend an Unrealistically Sophisticated Design?

Yes. AI may suggest a multi-site randomized trial, repeated longitudinal assessments, complex mediation models, and qualitative follow-up interviews without adequately considering the resources required.

Each component may be defensible in principle, but their combined implementation could exceed the researcher's available time, funding, or expertise.

Adding methodological complexity does not automatically strengthen a study. When procedures cannot be implemented consistently, complexity may create additional sources of error.

The problem of unrealistic AI research design recommendations is therefore distinct from whether the suggested methodology is theoretically appropriate.

Can AI Help Researchers Compare Feasibility Scenarios?

Yes. Scenario analysis is one of the more useful applications of AI-assisted planning.

Researchers may compare a preferred design under ideal conditions with alternatives that require fewer sites, shorter follow-up periods, or different data collection arrangements.

For example, a researcher may compare a multi-institutional randomized trial with a smaller single-institution trial or a carefully designed nonrandomized evaluation.

However, reducing scope may change the population, estimand, precision, or generalizability of the study.

AI should therefore explain what is lost or changed when a design is simplified, rather than presenting the alternatives as scientifically equivalent.

Should Researchers Use a Feasibility Score?

AI may offer a numerical feasibility score based on several criteria, but such scores should not be treated as validated measurements unless the underlying instrument and scoring procedure have appropriate evidence.

A score of 85% feasible, for example, has little scientific meaning if the weighting of recruitment, ethics, funding, and technical requirements is arbitrary.

A more transparent approach is to document each critical requirement, the evidence supporting its availability, the uncertainty remaining, and the consequences if it cannot be met.

Some constraints are decisive. A study that cannot obtain required ethical approval is not feasible merely because its budget and recruitment conditions are favorable.

When Should Researchers Conduct a Pilot or Feasibility Study?

When important implementation uncertainties cannot be resolved through existing information, preliminary empirical investigation may be appropriate.

For example, a researcher may be uncertain whether students will complete repeated assessments or whether instructors can deliver an AI-supported intervention consistently.

A pilot or feasibility study can provide direct evidence about these procedures.

Eldridge and colleagues (2016) emphasize that feasibility and pilot studies should investigate whether and how a future study can be conducted.

AI can help formulate preliminary objectives, but the decision to conduct a pilot or feasibility study should follow from the specific uncertainties that require investigation.

How Can You Conduct an AI-Assisted Feasibility Assessment?

The result should be a practical implementation rationale rather than an AI-generated declaration that the study is feasible.

When AI Recommends a Multi-Institutional Trial That Cannot Be Completed

The researcher asks AI to reassess the design using the actual constraints.

The researcher considers whether a smaller single-institution randomized study could address a narrower causal question, provided the sample size, allocation, intervention delivery, and outcome assessment remain appropriate.

Alternatively, a pilot randomized trial may be justified if important recruitment and implementation uncertainties remain unresolved.

The decision: The researcher revises the scope and research objectives to reflect what can realistically be implemented, rather than retaining an ambitious design whose essential requirements cannot be met.

The revised study may offer less generalizable evidence than the proposed multi-institutional trial, but it can still make a defensible contribution when its limitations are acknowledged.

The example demonstrates that methodological ambition and operational feasibility must be evaluated together.

Common Mistakes in AI-Generated Feasibility Assessments

Give AI Your Actual Constraints Before Asking Whether the Study Is Feasible

A useful feasibility assessment requires more than the research title and proposed methodology.

Provide the available research period, recruitment setting, institutional permissions, funding, technical infrastructure, and relevant methodological expertise. Ask AI to distinguish confirmed resources from assumptions requiring verification.

A Prompt for a Realistic Feasibility Assessment

The AI assessment should then be checked against actual institutional conditions and relevant methodological guidance.

When feasibility problems require substantial design changes, researchers should revisit the original study design decision and determine whether the revised approach still addresses the research problem.

A feasible research plan should be one the researcher can implement responsibly without weakening the intended inference beyond what the study can justify.

Before Accepting an AI Research Feasibility Assessment

Frequently Asked Questions About AI and Research Design Feasibility

A Research Design Must Be Possible to Implement, Not Merely Possible to Describe

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.