Researchers at East China Normal University have developed an interactive, AI-powered system designed to assess creative potential by mapping the cognitive processes students use to develop ideas rather than focusing solely on their final answers. Known as CREATIVA, this assessment tool utilizes large language models to capture the dynamic development of ideas, providing educators with a more comprehensive view of how students understand complex problems and respond to guidance. By prioritizing the creative process, the research team aims to identify potential that is often overlooked in traditional assessments, which typically emphasize the quantity and originality of final products.
Traditional creativity assessments have long relied on measuring the end results of students’ thinking, such as the number and novelty of their output. While these metrics provide useful performance snapshots, they offer limited insight into the underlying cognitive journey—the way students clarify problems, adjust their strategies, and refine their ideas after receiving support. To address these gaps, the CREATIVA system employs a process-dynamic approach that systematically records how students interact with scenarios and leverage AI-provided feedback to improve their solutions. This approach draws on the theoretical framework of the Zone of Proximal Creativity, which centers on the performance a student can achieve with appropriate guidance and how they respond to individual interventions throughout the creative task.
Built upon the educational large language model InnoSpark, the system operates across four distinct stages: problem discovery, problem identification, solution generation, and solution refinement. In a typical scenario involving an AI companion, students are first encouraged to identify significant challenges, choose a central problem to address, and approach it from multiple perspectives. The AI provides tailored background information and task guidance to ensure the student understands the scenario before engaging in the refinement process. During this latter stage, the system offers graduated prompts that challenge students to consider the feasibility of their proposals, identify potential risks, and explore areas for improvement. This structured interaction allows the assessment to capture real-time adaptations in the student’s problem-solving approach.
To evaluate the system’s efficacy, researchers conducted an initial pilot study with 29 undergraduate students enrolled in a psychology course. Participants completed the CREATIVA assessment alongside established measures, including the Alternative Uses Task and Realistic Presented Problems. Five independent graduate raters assessed participant responses using the established framework, finding a high degree of consistency across the 12 scoring dimensions, with consistency coefficients ranging from 0.80 to 0.94. Furthermore, the study identified meaningful associations between specific CREATIVA indicators and scores on traditional creativity measures, providing preliminary psychometric support for the system’s validity and suggesting that the platform captures aspects of creative performance related to those recognized by established tools.
Beyond the undergraduate pilot, the research team implemented CREATIVA with a broader demographic, including elementary, junior high, and senior high school students. By adapting task scenarios and guidance levels to fit curricular content and background knowledge, the team observed that some students who struggled during the initial problem discovery stage were nonetheless capable of submitting high-quality solutions after receiving AI support. This finding suggests that responsiveness to guidance may serve as a crucial indicator of creative potential, revealing capabilities that might otherwise remain invisible if an assessment only captures initial performance or finished work.
Despite these encouraging preliminary results, the researchers emphasize that the system is in an early stage of development. The current pilots do not establish that the CREATIVA system leads to lasting, long-term improvements in creativity, nor has it been broadly validated across diverse educational settings or subjects. Future research is required to determine the system’s applicability in different contexts and to establish how its assessment data can be most effectively utilized to inform classroom instructional support and talent identification processes.
The researchers hope that by moving toward this process-oriented assessment, educators can better understand the full spectrum of learners’ abilities.
Subject of Research: AI-driven assessment of creative process in educational settings.
Article Title: Beyond the final answer: new AI system tracks how students develop creative ideas
Article References: Beyond the final answer: new AI system tracks how students develop creative ideas. Original publication (Bibliographic details unavailable.)
Image Credits: AI Generated
DOI: Not provided
Keywords: Artificial Intelligence, Creative Assessment, Education Technology, InnoSpark, ECNU, Problem Solving, Developmental Potential, Process-Dynamic Approach, Zone of Proximal Creativity, Educational Psychology, Creative Thinking, Student Evaluation, Human-AI Collaboration, Psychometric Validation, Creative Potential
Cite Scienmag News
Courtney Benton. (November 18, 2026). AI System Evaluates Creative Potential By Tracking Student Problem-Solving Processes. Scienmag. https://scienmag.com/ai-system-evaluates-creative-potential-by-tracking-student-problem-solving-processes/
Courtney Benton. "AI System Evaluates Creative Potential By Tracking Student Problem-Solving Processes." Scienmag, 18 November 2026, https://scienmag.com/ai-system-evaluates-creative-potential-by-tracking-student-problem-solving-processes/. Accessed 18 November 2026.
Courtney Benton. "AI System Evaluates Creative Potential By Tracking Student Problem-Solving Processes." Scienmag. November 18, 2026. https://scienmag.com/ai-system-evaluates-creative-potential-by-tracking-student-problem-solving-processes/








