Charting the Future of AI in Project-Based Learning: A Co-Design Exploration with Students

• AI tools can assist learning tasks while creating concern that assessments may measure AI capabilities rather than students’ acquired and applied skills.In PBL, this concern is especially relevant because students produce artifacts that commonly serve as learning indicators.Source: §1 · pp. 1–2
• Reports documenting students’ learning processes and AI assistance could expose higher-order thinking and clarify how students’ efforts contribute to project outcomes.The paper identifies students’ desired ways of reporting AI involvement as an unresolved assessment question.Source: §1 · p. 2
• The study uses a co-design workshop in which students imagine an alternative present with advanced AI capabilities and design reports documenting their AI use in PBL.The design responds to immature adoption, limited access, and evolving AI services by enabling speculative exploration.Source: §1 · p. 2
• The workshop links AI-involved project journeys, future ideal-student traits, and AI usage report design to explore how AI interaction data might support assessment.Participants first reconsider project processes, then identify valued traits, and finally connect AI usage data to those traits.Source: §1 · p. 2
• Across seven three-hour workshops with 18 college students, the study found varied AI-use scenarios, multiple analysis approaches, attitude-dependent reporting preferences, and concerns about ambiguous interpretation.The qualitative findings are not statistically generalizable, but they provide detailed insights into students’ views on AI-enhanced PBL.Source: §1 · p. 2
• The paper contributes future AI-use scenarios, education-goal transformations, analysis possibilities, and research opportunities concerning student-AI interaction and sensemaking.Source: §1 · p. 2

2.1 Project-Based Learning

• PBL emphasizes authentic, extended projects and higher-order learning, but artifact-focused and traditional assessments can overlook the learning process.Complementary presentations, journals, portfolios, and self-reflection are used to make that process more visible.Source: §2.1 · pp. 2–3
• Technology supports information processing and higher-order thinking in PBL, yet AI’s human-like capabilities raise distinct questions about appropriate roles and future assessment.The paper therefore examines how AI may reshape PBL learning goals and assessment.Source: §2.1 · p. 3

2.2 AI Tools to Support Learning Tasks & Generative AI (GenAI)

• AI tools support learning through feedback, scaffolding, logistics, and task collaboration, while students increasingly access general-purpose tools for brainstorming, data work, writing, and presentations.GenAI tools such as ChatGPT and Midjourney have made assistance broadly accessible in student learning.Source: §2.2 · p. 3
• GenAI adoption raises concerns about cheating, indistinguishable authorship, inaccurate or biased information, and institutional restrictions.These concerns complicate responsible adoption and motivate greater transparency in students’ AI use.Source: §2.2 · p. 3
• Researchers increasingly anticipate AI becoming normal in learning, shifting assessment toward process and AI literacy while requiring transparency to understand critical thinking and problem-solving.The paper addresses the stated lack of research on collecting and analyzing this transparency in AI-enhanced education.Source: §2.2 · p. 3

2.3 Tracking and Sensemaking of Learning Process

• Learning-process tracking spans platform analytics, instructor observations, and student self-tracking, but documentation can create tension between creating and recording.Source: §2.3 · pp. 3–4
• This study focuses on extended, non-classroom PBL where powerful AI supports learning and students’ AI interaction data are analyzed to understand learning and evaluate outcomes.Unlike prior AI-usage studies centered on separate research questions, it investigates how students view the educational use of those data.Source: §2.3 · p. 4

3.1 Participants

• The study recruited 18 students from four East Asian higher-education institutions, requiring PBL and AI-tool experience and reaching theoretical saturation.Participants included undergraduates and graduates across varied majors; all had used ChatGPT and many had used AI in course projects.Source: §3.1 · pp. 4–5

3.2 Workshop Setup

• Seven workshops lasted about three hours and were ultimately conducted virtually using videoconferencing and Miro, with sessions split to manage fatigue.Source: §3.2 · p. 5

3.3 Workshop Design

• The workshop framework assumed AI would transform learning processes and assessment, so students designed ways to analyze and report AI usage as assessment material.The design was finalized through four pilot rounds and drew on speculative design methods.Source: §3.3 · p. 5, §3.3.1 · p. 5

3.3.1 Activity 1: AI-involved Project Journey Speculation.

• Activity 1 grounded speculation in students’ past PBL experiences, then reimagined them with AI ten years more advanced and mapped possible AI-supported project journeys.Learner journey templates, AI capability cards, and peer discussion supported the activity.Source: §3.3.1 · pp. 5–6

3.3.2 Activity 2: Imagine the Ideal Student.

• Activity 2 examined traits challenged by AI and brainstormed future traits, including AI application abilities, while considering how instructors, students, and employers might value them.Participants used an adapted persona-building approach to imagine ideal future students.Source: §3.3.2 · pp. 6–8

3.3.3 Activity 3: AI Usage Report Design.

• Activity 3 mapped AI interaction data to valued traits and designed hypothetical reports showing what a magic project studio might capture for instructors, self-reflection, or job seeking.Participants considered AI type, project stage, inputs, customization, and handling of outputs, then explained how designs supported assessment.Source: §3.3.3 · p. 8

3.4 Data Analysis

• Three researchers used inductive thematic analysis with concurrent analysis, source triangulation, investigator discussion, and participant member checking.Eleven participants returned annotated reports used to refine the analysis.Source: §3.4 · pp. 8–9

• Six primary themes emerged from the analysis.Source: §4 · p. 9
• The first three themes align with the workshop’s three activities.Source: §4 · p. 9
• The latter three themes emerged from participants’ comments and discussions.Source: §4 · p. 9

• Over 100 desired AI usages from seven workshops with 18 participants clustered into six purposes for supporting PBL.The purposes were automating repetitive tasks, supporting divergent thinking, selecting among alternatives, drafting or implementing solutions, providing feedback, and guiding learning.Source: §4.1 · p. 9
• Students wanted AI to automate time-consuming work while stimulating divergent thinking through varied ideas and filtering.Source: §4.1 · p. 9
• Students envisioned AI comparing alternatives, including algorithms or literature, to help select effective ideas.Source: §4.1 · p. 9
• Students wanted AI to turn their foundational solution concepts into code, prototypes, or visualizations.Source: §4.1 · p. 9
• Students also envisioned AI evaluating solution effectiveness and feasibility while teaching concepts or assessing readiness to undertake projects.Source: §4.1 · p. 9

4.2 Students’ Envisioned Future Assessment Transformation

• Most participants viewed traditional artifact-based assessment as inadequate for creative thinking and efficiency, although some retained artifacts’ importance because AI may lack contextual precision.Source: §4.2.1 · p. 9, §4.2 · p. 9

4.2.1 Old Traits Made Different by AI.

• Participants disagreed about whether logical and critical thinking would remain essential as AI’s reasoning capabilities grow.Some considered these skills less critical, while P17 linked critical thinking to deciding how to question and direct AI.Source: §4.2.1 · p. 9
• Participants identified AI-related traits they considered important because of AI usage, beyond traits currently assessed.Source: §4.2.2 · p. 9

4.2.2 New Traits Needed Due to Students’ AI Adoption.

• The envisioned ideal student uses AI purposefully, communicates clearly, retains project direction, and treats AI as a supporting actor.Source: §4.2.2 · p. 9
• Participants also emphasized symbiotic student-AI learning and judgment in selecting high-quality options as AI becomes more capable.Source: §4.2.2 · p. 9

4.3 Students’ Designs of Reporting of AI Usages

• The proposed categories open discussion about AI-usage analysis rather than claiming exclusive or representative coverage.Not every usage analysis had a corresponding data framing, possibly because of participants’ data and visualization literacy.Source: §4.3 · p. 10

4.3.1 Task Allocation between Students and AI.

• Students proposed distinguishing human- and AI-performed tasks because task allocation indicates which skills students practiced or demonstrated.P16’s Gantt-style design encoded task ownership, time spent, and required creativity.Source: §4.3.1 · p. 10
• Participants also wanted to show complementary human-AI roles, with AI handling repetitive work while students retained responsibility for creative tasks.Some students nevertheless viewed their work with AI as a team effort rather than a clean division of labor.Source: §4.3.1 · pp. 10–11

4.3.2 Engagement of AI in the Project.

• Students proposed measuring AI’s project engagement through contribution percentages, usage frequency, generated word counts, solved problems, and artifact annotations.They questioned whether simple metrics capture human input, and suggested distinguishing direct AI output, edited output, and human content inspired by AI.Source: §4.3.2 · pp. 11–12

4.3.3 Effectiveness of Students-AI Interaction.

• Students linked effective AI use to question design, iterative prompting, time and question counts, peer comparisons, and changes in AI behavior caused by continued input.They also proposed comparing original and AI-improved work to assess AI mastery and support reflection.Source: §4.3.3 · p. 12

4.3.4 The Process of Students Incorporating AI’s Suggestions into the Project.

• Students viewed how they filter, question, synthesize, and iteratively combine AI suggestions as evidence of critical thinking, creativity, leadership, and mutual enhancement.Proposed framings included funnel charts, questioning-frequency charts, multi-source synthesis, and bidirectional refinement diagrams.Source: §4.3.3 · p. 12

4.3.5 Quantifying Student Development through Human-AI Interac-

• Participants proposed tracking student development through AI interactions, including growing AI-use skill and possible negative effects of relying on AI for tasks.Some designs deducted skill scores for AI-completed work, while others increased scores for manual work or rich interaction.Source: §4.3.5 · pp. 12–13

4.3.6 AI Impact on Student-Student Collaboration.

• Students wanted AI-use analysis to examine whether AI changes human collaboration by reducing teammate communication or creating conflicts over acceptable uses.They suggested contrasting student-student with student-AI communication, but did not develop a framing design for collaboration impacts.Source: §4.3.6 · p. 13

• Participants identified three AI roles—tool, teammate, and expert—that shaped their preferred ways of analyzing AI use.Tool-oriented students emphasized human leadership; teammate-oriented students prioritized collaborative outcomes; expert-oriented students emphasized critical evaluation of AI advice.Source: §4.4 · p. 13
• Tool-oriented participants wanted students to retain project direction and limit AI to supportive or trivial tasks.Source: §4.4 · p. 13
• Teammate-oriented participants prioritized effective joint outcomes and viewed separating human and AI contributions as unnecessary or undesirable.Source: §4.4 · p. 13
• Expert-oriented participants required reports to show how cautiously students integrated AI suggestions that could be biased, subjective, or misleading.Source: §4.4 · p. 13

• For instructor assessment, participants favored holistic, in-depth analysis of AI use to inform learning assessment.Source: §4.5 · p. 13
• For job-seeking evaluations, participants emphasized demonstrating efficiency in leveraging AI technologies.Source: §4.5 · p. 13
• For self-assessment, participants emphasized reflection, including distinguishing learning-oriented interactions and integrating personal measures such as emotions or heart rate.Source: §4.5 · p. 13

• Participants worried that interaction-based assessment could unfairly penalize students who rely on strong AI suggestions or reward superficial edits as critical thinking.Source: §4.6 · p. 13
• AI’s limited sensing and understanding could produce unwarranted credit when students modify AI-generated content using contextual knowledge.Source: §4.6 · p. 13
• Students might alter their learning behavior or fabricate a more attractive AI-usage report to satisfy instructors’ evaluation criteria.Source: §4.6 · p. 14

5 DISCUSSION

• The study suggests AI-interaction data can augment artifact-based assessment and reveal higher-order thinking and effective AI use, but fairness and hacking remain concerns.Source: §5 · p. 14

5.1.1 How do students’ perceptions of AI roles influence educational interactions?

• Students’ beliefs about AI as a tool, teammate, or expert influence AI use and analysis, while mismatched beliefs may create conflicts or ineffective pedagogy.Source: §5.1.1 · p. 14

5.1.2 How can we tailor AI for students to use in PBL?.

• General-purpose AI may threaten assessment fairness, motivating student-centered tools that personalize assistance or scaffold responses to students’ skill levels.Source: §5.1.1 · p. 14

5.1.3 How can we support self-regulated learning in AI-enhanced environments?

• Because AI over-reliance may impede self-assessment and independent strategy adjustment, future research should examine interaction monitoring, documentation, and learning analytics for supporting self-regulated learning.Source: §5.1.3 · p. 14

5.1.4 How should education practitioners balance the goals of ef-

• Delegating tasks to AI may improve effective use while bypassing the active construction of knowledge that underpins PBL, creating a need to balance efficiency with active learning.Source: §5.1.4 · pp. 14–15

5.1.5 What are the impacts of AI on communication in education?

• Replacing instructors with AI may misalign students’ learning with curriculum and teaching plans, while also affecting student-student communication.Source: §5.1.5 · p. 15

5.2.1 How can we support the collection of data around students use of AI?.

• Students’ AI-use data should connect usage contexts, students’ thoughts and actions, and AI outputs to corresponding solution components.Analysis may distinguish task automation from feedback use and trace how students incorporate AI suggestions into project artifacts.Source: §5.2.1 · p. 15
• Collecting students’ motivations, thoughts, and transformations from AI outputs into solutions is non-trivial and requires documentation tools and provenance research.The paper identifies student input, interactive documentation, and information-provenance methods as directions for future work.Source: §5.2.1 · p. 15

5.2.2 How can we make sense of AI’s contribution based on students-AI interaction data?

• Students viewed AI-contribution assessment as essential, but evolving human-AI interaction may make the parties’ contributions difficult to disentangle.Future analyses should combine qualitative and quantitative interaction evidence to examine whether students remain in leading positions rather than reduce contribution to a single percentage.Source: §5.2.2 · pp. 15–16

5.2.3 How can we support sensemaking of students’ use of AI from multiple perspectives?

• Different students assign different meanings to the same AI-use measures, so practitioners need multi-perspective sensemaking rather than assuming one correct interpretation.Shorter interactions may signal efficient mastery or limited engagement, while larger or smaller AI-contribution shares may reflect different educational values.Source: §5.2.2 · p. 16
• As AI becomes more personalized, context-sensitive, and creative, students may support transferring critical-thinking, decision-making, and creative roles to AI, conflicting with educational ideals that retain those roles for humans.The paper does not judge which values are more correct and calls for fair, comprehensive interpretation involving students themselves.Source: §5.2.2 · p. 16

5.2.4 How can we motivate students to document their use of AI and report it honestly?

• Students may strategically misreport AI use when reports affect grades, so documentation should support learning and be assessed alongside multiple factors rather than function solely as a grading tool.The paper suggests communicating the benefits of honest documentation and using reports to improve instruction.Source: §5.2.4 · p. 16

5.3 Limitation & Future Work

• The qualitative workshop findings cannot be generalized statistically to broader institutions, populations, courses, or extended periods.Future research should test analytic designs, anticipated AI roles, and scenario effects with larger samples and longer studies.Source: §5.3 · pp. 16–17
• The study’s three-hour, hypothetical, student-centered workshops differ from weeks- or months-long PBL with hands-on AI use and educator perspectives.Longitudinal studies in actual PBL settings could provide more nuanced evidence about AI interaction, usage-data interpretation, and assessment.Source: §5.3 · p. 17

6 CONCLUSION

• Students envisioned AI-usage data as material for assessing skills, decision-making, ethical awareness, and contributions in PBL.The study connects these possibilities to future education systems, documentation tools, and learning analytics.Source: §6 · p. 17
• Students differed in whether they viewed AI as an ability-augmenting tool, teammate, or expert, and these beliefs shaped their use and reporting of AI.Source: §6 · p. 17