Can Generative AI Identify Double-Barreled Survey Questions?

Can AI Detect Survey Questions That Ask Two Things at Once?

You are reviewing a research questionnaire and encounter the statement: "Generative AI helps me prepare lessons and assess students effectively."

At first glance, the item appears straightforward. Both activities involve teaching, and respondents can indicate their level of agreement. However, a teacher might believe that AI improves lesson preparation while finding it unsuitable for student assessment. What should that teacher answer?

This is the problem with double-barreled questions. They combine multiple judgments into one response, potentially making the resulting data difficult to interpret. Generative AI can help identify these items, but detecting them requires more than searching for sentences containing the word "and."

Yes, AI Can Identify Potentially Double-Barreled Questions

How to Recognize and Correct Double-Barreled Survey Questions

What Makes a Survey Question Double-Barreled?

A double-barreled question asks respondents to evaluate two or more distinguishable matters using a single answer. The problem arises when the combined components do not necessarily produce the same response.

Consider the question: "How satisfied are you with your university's internet connection and technical support?"

A respondent may be satisfied with the internet connection but dissatisfied with technical support. Selecting a single satisfaction rating would obscure this distinction.

The question can be separated into two items:

• How satisfied are you with your university's internet connection?
• How satisfied are you with your university's technical support?

Each revised question now asks for one evaluation. The researcher can analyze satisfaction with the two services separately rather than interpreting an ambiguous combined response.

Double-barreled questions are commonly discussed in survey methodology because they can undermine response clarity and complicate interpretation.

The Most Useful Test: Could the Respondent Give Different Answers?

A practical way to identify a double-barreled question is to ask whether a respondent could reasonably agree with one component while disagreeing with another.

For example, "I am confident in designing research instruments and analyzing quantitative data" combines two distinct research capabilities. A researcher might feel highly confident designing questionnaires but less confident performing statistical analysis.

If the item uses a single agreement scale, the respondent must somehow combine these judgments. The resulting answer may not clearly represent either capability.

This test is more useful than simply checking for conjunctions. A question may contain "and" without requiring separate judgments, while another question may combine distinct concepts without using the word at all.

What Types of Double-Barreled Questions Can AI Detect?

AI can examine sentence structure and identify questions that combine different actions, attributes, experiences, or evaluations.

These examples show why a question can appear conceptually coherent while still requiring multiple judgments. Related concepts are not necessarily interchangeable.

AI can be particularly useful for reviewing long questionnaires where such combinations are easy to overlook during manual editing.

Not Every Question Containing "And" Is Double-Barreled

AI may incorrectly flag questions simply because they contain conjunctions. Yet the grammatical presence of "and" does not automatically indicate a measurement problem.

Consider the statement: "I can identify the independent and dependent variables in a research study."

Here, identifying both types of variables may reasonably be treated as one specified academic task, depending on the intended construct. The conjunction describes the task rather than necessarily introducing two separate evaluations.

However, if the researcher specifically wants to distinguish students' abilities to identify independent variables from their abilities to identify dependent variables, separate items may be more appropriate.

The decision therefore depends on the intended measurement. AI should identify potentially separable components, while the researcher determines whether those distinctions matter for the study.

Can a Question Be Double-Barreled Without Using "And"?

Yes. Some questions combine multiple judgments through broad phrases, compound descriptions, or concepts that are treated as though they were a single attribute.

For example, "How would you rate the overall quality of your university's digital infrastructure?" could be a legitimate global evaluation. But if the researcher intends to measure internet reliability, software availability, and technical support separately, the question would be insufficient for that purpose.

Similarly, asking respondents whether an AI tool is "efficient, accurate, and trustworthy" may invite them to evaluate three characteristics together even though their experiences with each may differ.

AI can help identify these conceptual combinations, but the researcher must distinguish between a deliberately broad overall assessment and an item that unintentionally combines separate constructs.

How Should You Prompt AI to Detect Double-Barreled Questions?

A useful prompt asks AI to identify independently answerable components, explain the problem, and suggest revisions without unnecessarily splitting legitimate single-focus questions.

This prompt encourages a concept-based review rather than a purely grammatical one. It also helps the researcher examine whether the proposed revisions improve interpretability.

AI can assist in improving questionnaire item wording, but a clearer sentence is not necessarily a better measure unless it remains aligned with the intended construct.

Should Every Double-Barreled Question Be Split Into Two Items?

Splitting is often appropriate when the combined components represent different attributes or activities that the researcher needs to measure separately. However, it is not the only possible solution.

Sometimes the original question is unnecessarily broad, and the researcher needs only one component. In that situation, removing the irrelevant component may be preferable to creating an additional item.

For example, if a study measures teachers' perceived usefulness of AI for lesson preparation, the item "AI improves my lesson preparation and student assessment" may be revised to focus only on lesson preparation. Adding a separate assessment item would be unnecessary unless student assessment also belongs within the intended measurement domain.

In other cases, a global judgment is deliberately intended. A researcher might ask respondents to evaluate their overall satisfaction with a university service. Such an item may be acceptable as a broad assessment, provided the researcher does not later interpret it as a precise measure of each component.

The choice should follow the research objective, not a mechanical rule that every compound sentence must be divided.

What Happens When Double-Barreled Items Are Used in Likert Scales?

Double-barreled statements are particularly problematic in multi-item scales because they can obscure what a response represents.

Suppose an instrument measures research self-efficacy and includes the statement "I can formulate research questions and analyze quantitative data."

A respondent who feels confident about formulating questions but not analyzing data may select the midpoint. Another respondent with moderate confidence in both tasks may choose the same response.

Although the numerical answers are identical, the underlying judgments differ. This ambiguity complicates interpretation of the item and may affect how it relates to other scale items.

When using AI to generate Likert-scale items, researchers should therefore evaluate whether each statement represents a sufficiently coherent and interpretable target.

Removing double-barreled wording is part of item development, not proof of validity. A revised item still requires appropriate content evaluation and, when relevant, empirical investigation of its measurement properties.

Can AI Detect Every Double-Barreled Question Reliably?

No. AI may overlook subtle conceptual combinations or flag legitimate compound expressions as problematic. Its judgment can also vary depending on how the task is prompted and how much information is supplied about the construct.

For instance, an AI system reviewing an item about "research planning and implementation" may treat the phrase as a single activity, while a researcher familiar with the framework may recognize two theoretically distinct dimensions.

Conversely, AI may recommend splitting a question that intentionally measures a specific integrated task.

Researchers should use AI-generated flags as review candidates rather than definitive classifications. Expert assessment and respondent pretesting can help determine whether the wording is understandable and whether separate judgments are required.

Importantly, cognitive interviewing and pilot testing provide information about actual response processes that an AI-only review cannot establish.

Using AI to Separate Two Teaching Activities in One Question

The revisions allow respondents to distinguish the usefulness of AI for two activities. However, splitting the original statement does not automatically establish that both revised items should belong in the same scale.

The researcher must still decide whether the activities adequately represent the construct and whether their responses support the intended score interpretation.

Common Mistakes When Identifying Double-Barreled Questions

Use AI to Identify Multiple Judgments, Then Decide What Needs Measuring

The most useful way to employ AI is to ask it to identify questions where respondents could reasonably hold different views about the components. Request the specific phrases involved and possible revisions, but retain control over whether the item should be divided, narrowed, or preserved.

A double-barreled question is not merely a writing problem. It can also reveal uncertainty about what the researcher intends to measure. If a question combines two constructs, the underlying issue may be conceptual rather than grammatical.

After revising the items, check their response options and evaluate whether the changes affect the questionnaire's intended content. AI can accelerate this review, but it cannot independently establish that the revised questions will function as intended.

Before Accepting AI-Suggested Revisions to Double-Barreled Questions

Frequently Asked Questions About Double-Barreled Survey Questions

AI Can Detect Multiple Questions Hidden Inside One Item

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.