Creativity
Conditions
Keywords
Artificial Intelligence, Generative AI, ChatGPT, Human-AI Collaboration, Problem Solving, Teamwork, Undergraduate Students, Ideation, Brainstorming, Creativity
Brief summary
This study aims to examine how the timing of artificial intelligence (AI) integration during group brainstorming activities affects the creativity and ideation of participants. Undergraduate student participants will be randomly allocated into teams of four to complete a structured 12-minute brainstorming task. The study utilizes a three-group experimental design to compare three conditions: no AI access, AI access from the outset of the session, and AI access introduced halfway through the session. The standardized generative AI tool used is ChatGPT Pro. The creative output produced by each group will be evaluated by a panel of independent expert judges using the Consensual Assessment Technique (CAT).
Detailed description
Artificial intelligence (AI) tools are increasingly used in academic and professional environments, yet little is known about how the timing of AI's introduction during a collaborative brainstorming process shapes the creativity of the ideas produced. While recent literature suggests that human-AI collaboration can increase the likelihood of successful idea generation when AI is present throughout an entire session, the specific effect of integration timing remains a gap in the literature. This study aims to provide early data on whether AI functions more effectively as an early collaborator, a later resource, or whether its absence supports more creative thinking. The study recruits undergraduate students enrolled in a summer research program, who are then randomly assigned to teams of four using a digital randomizer. Over the course of nine sessions, these teams participate in structured, 12-minute collaborative brainstorming tasks focused on designing strategies to address common university student challenges. To mitigate the influence of individual baseline characteristics on the group output, participants are re-randomized into new teams for each separate session. To standardize the technological intervention and eliminate technical confounders, the study utilizes ChatGPT Pro (GPT-4o architecture) on pre-set devices. At the conclusion of the timed task, a designated team scribe submits the group's collaboratively generated idea list and their selected best idea via a secure, digital case report form. Following data collection, all session outputs are immediately de-identified and assigned neutral identifiers. A panel of independent expert judges, blinded to the condition assignments, evaluate the output of each group.
Interventions
BEHAVIORALArtificial Intelligence First (Early Integration)
The introduction of ChatGPT Pro (GPT-4o architecture) at the beginning (minute 0) of a 12-minute group brainstorming session. A single prompt containing the scenario is submitted to generate an initial set of ideas, with no further back-and-forth dialogue permitted. This single artificial intelligence output serves as the baseline for the team's subsequent human ideation.
BEHAVIORALArtificial Intelligence Last (Late Integration)
The introduction of ChatGPT Pro (GPT-4o architecture) at the midpoint (minute 6) of a 12-minute group brainstorming session. A single prompt containing both the scenario and the team's already-generated human ideas is submitted, with no further back-and-forth dialogue permitted. This single artificial intelligence output is used to guide the team's human ideation for the remaining 6 minutes.
Sponsors
McMaster University
Study design
Allocation
RANDOMIZED
Intervention model
PARALLEL
Primary purpose
OTHER
Masking
SINGLE (Outcomes Assessor)
Intervention model description
The study uses a three-group experimental design where participants are allocated into teams of four and randomly assigned to one of three parallel experimental conditions: No AI, AI First, and AI Last. This design allows for the direct comparison of how the timing and availability of artificial intelligence access shapes creative output during a structured brainstorming task under otherwise identical conditions.
Eligibility
Sex/Gender
ALL
Healthy volunteers
Yes
Inclusion criteria
* Enrolled as a student in the designated summer research program * Able to provide voluntary informed consent * Proficient in reading and communicating in English (required to engage with the scenarios and the artificial intelligence tool)
Exclusion criteria
-Not enrolled as an undergraduate student
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Creativity Score of the Selected Best Idea | Evaluated cross-sectionally upon the conclusion of each 12-minute session, across the 9 study sessions. | The primary outcome is the mean expert creativity rating assigned to each team's top idea, across each of the 9 sessions, as assessed by independent judges using the Consensual Assessment Technique (CAT). Higher average scores denote a more creative output, while lower scores denote a less creative output. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Overall Creativity Score of Full Idea List | Evaluated cross-sectionally upon the conclusion of each 12-minute session, across the 9 study sessions. | The collective creative quality of the entire list of ideas generated by the team during the session, evaluated by independent expert judges using the Consensual Assessment Technique (CAT). |
| Number of Ideas Generated During the 12-Minute Session | Evaluated cross-sectionally upon the conclusion of each 12-minute session, across the 9 study sessions. | The total number of ideas generated (fluency) will be reported with the explicit methodological limitation that AI generation speed inherently outpaces human typing speed. Fluency will be analysed purely observationally, as the 12-minute constraint does not control for the mechanical advantage of the AI tool. |
| Qualitative Resemblance of the Team's Output to Typical AI-Generated Responses | Evaluated cross-sectionally upon the conclusion of each 12-minute session, across the 9 study sessions. | This will be subjectively assessed by the expert judges on a 1-5 scale (anchored from 'strongly AI-typical' to 'strongly human'), independent of the CAT creativity evaluation. |
Countries
Canada
Contacts
CONTACTAlexandra Leone, BSc
CONTACTKaya Bhandari
PRINCIPAL_INVESTIGATORMohit Bhandari, MD, PhD, FRCSC
McMaster University
Outcome results
None listed