This analysis is performed from the perspective of a Senior Instructional Design and Educational Assessment Specialist, focused on research methodology, survey instrument construction, and data integrity as applied to educational technology studies.
Abstract:
This instructional video details the critical process of translating a pre-existing research questionnaire into a functional and methodologically sound Google Form survey instrument. The instructor emphasizes the need for rigorous validation of AI-generated content, strict adherence to research protocols (e.g., limiting scope to surveys, not experiments), and careful construction of survey elements to mitigate common data collection errors. Key areas addressed include refining demographic sections, avoiding leading questions by separating variables, ensuring appropriate scaling (e.g., preferring 6-point linear scales over 5-point Likert scales or rating grids for complex constructs), and utilizing Google Form's response validation features for quantitative data integrity (e.g., grade entry). The overall objective is to establish a robust data collection mechanism that minimizes post-collection cleaning and encoding overhead, ensuring data reliability for subsequent statistical analysis linking AI tool usage to mathematical competence.
Reviewer Recommendation:
This material is essential for Graduate-Level Research Assistants, Undergraduate Thesis/Capstone Students in Applied Sciences or Education, and Instructional Technologists tasked with designing primary data collection instruments for quantitative studies involving technology integration.
Summary of Google Form Construction Methodology
0:00 General Scope & AI Acknowledgment: The discussion centers on converting questionnaires into Google Forms for survey administration. A strict disclaimer is issued that all AI-generated content (like the sample questionnaire) must be manually verified and edited, as it is not automatically correct.
0:58 Study Design Focus: The research scope is explicitly limited to surveys, excluding experimental studies. The core research problem is investigating the effect of AI-assisted confidence (Independent Variable) on mathematical competence (Dependent Variable), noting the observed issue of overconfidence due to AI reliance.
2:47 Demographic Data Control: Demographic profiling (age, sex, program) is deemed optional but useful for potential correlation testing. The instructor stresses standardization by using explicit choices rather than open text fields (e.g., to prevent "BS Nursing" vs. "Nursing" entries). Specific fields (like Year Level) are removed if the cohort is homogeneous (e.g., all first-year students).
4:35 Handling Missing Data: Clear instructions are provided on handling data points like preliminary grades: respondents must leave the field blank if the data is unavailable; entering 'zero' will incorrectly influence computations.
7:15 Question Refinement (Bias Mitigation): Questions linking the Independent Variable (AI use) and Dependent Variable (Math competence) directly in the prompt phrasing must be revised. Such phrasing introduces immediate bias (e.g., "I'm confident because of AI") by cueing the respondent to the expected relationship.
9:50 Survey Sectioning and Formatting: The form should be organized into logical sections (e.g., Part 1: Demographics, Part 2: AI Confidence Scale).
17:38 Informed Consent Protocol: The consent section must be manually typed, with placeholders replaced by the researcher's information. Crucially, respondents must not use the "I disagree" option, as this invalidates their entire subsequent response set.
20:49 Response Validation for Numerical Integrity: Response validation is a key technique used for numeric fields (e.g., grades). For instance, grades scored out of 50 must be validated to ensure entries are numbers and fall within a logical range (e.g., $\geq 50$ and $\leq 100$), preventing erroneous data entry.
25:19 Ordinal Data Ordering: Options for ordinal data (e.g., frequency scales like "Never" to "Always") must be arranged logically by rank, not alphabetically.
26:40 Checkbox vs. Single Choice: Checkboxes (allowing multiple selections) are discouraged for variables where only one response is expected (e.g., primary device used), as this leads to complex data categorization in the output sheet.
28:02 Likert Scale Selection: The instructor explicitly prefers a 6-point scale over the common 5-point scale (which includes a neutral point) to force a directional choice, noting that validated scales should not be altered.
38:24 Data Output Preference (Numericalization): The goal is to structure scales (e.g., Agree/Disagree) using numerical options (1-6) in the form, as this immediately yields quantifiable data in the linked Google Sheet, simplifying computation later. Using descriptive words (e.g., "Strongly Agree") as choices results in non-numeric text columns upon export.
41:03 Previewing and Publishing: Forms must be previewed by the researcher, who should attempt to answer the survey, to catch sequencing issues or conditional logic failures. The form must then be explicitly published and set to accept responses.
46:41 Data Export to Sheets: The final, crucial step is linking the form to a Google Sheet to retrieve the structured data. Checkbox responses, which appear as concatenated strings in separate columns, will require further data wrangling in subsequent analytical lessons.