Exploration vs. Fixation: Scaffolding Divergent and Convergent Thinking for Human-AI Co-Creation with Generative Models

Generative AI has begun to democratize creative work, enabling novices to produce complex artifacts such as code, images, and videos. However, in practice, existing interaction paradigms often fail to support divergent exploration: users tend to converge too quickly on early ``good enough'' results and struggle to move beyond them, leading to premature convergence and design fixation that constrains their creative potential. To address this, we propose a structured, process-oriented human-AI co-creation paradigm including divergent and convergent thinking stages, grounded in Wallas's model of creativity. To avoid design fixation, our paradigm scaffolds both high-level exploration of conceptual ideas in the early divergent thinking phase and low-level exploration of variations in the later convergent thinking phrase. We instantiate this paradigm in HAIExplore, an image co-creation system that (i) scaffolds divergent thinking through a dedicated brainstorming stage for exploring high-level ideas in a conceptual space, and (ii) scaffolds convergent refinement through an interface that externalizes users' refinement intentions as interpretable parameters and options, making the refinement process more controllable and easier to explore. We report on a within-subjects study comparing HAIExplore with a widely used linear chat interface (ChatGPT) for creative image generation. Our findings show that explicitly scaffolding the creative process into brainstorming and refinement stages can mitigate design fixation, improve perceived controllability and alignment with users' intentions, and better support the non-linear nature of creative work. We conclude with design implications for future creativity support tools and human-AI co-creation workflows.

Automating Human Tutor-Style Programming Feedback: Leveraging GPT-4 Tutor Model for Hint Generation and GPT-3.5 Student Model for Hint Validation

Generative AI and large language models hold great promise in enhancing programming education by automatically generating individualized feedback for students. We investigate the role of generative AI models in providing human tutor-style programming hints to help students resolve errors in their buggy programs. Recent works have benchmarked state-of-the-art models for various feedback generation scenarios; however, their overall quality is still inferior to human tutors and not yet ready for real-world deployment. In this paper, we seek to push the limits of generative AI models toward providing high-quality programming hints and develop a novel technique, GPT4Hints-GPT3.5Val. As a first step, our technique leverages GPT-4 as a ``tutor'' model to generate hints -- it boosts the generative quality by using symbolic information of failing test cases and fixes in prompts. As a next step, our technique leverages GPT-3.5, a weaker model, as a ``student'' model to further validate the hint quality -- it performs an automatic quality validation by simulating the potential utility of providing this feedback. We show the efficacy of our technique via extensive evaluation using three real-world datasets of Python programs covering a variety of concepts ranging from basic algorithms to regular expressions and data analysis using pandas library.

Generative AI for Programming Education: Benchmarking ChatGPT, GPT-4, and Human Tutors

Generative AI and large language models hold great promise in enhancing computing education by powering next-generation educational technologies for introductory programming. Recent works have studied these models for different scenarios relevant to programming education; however, these works are limited for several reasons, as they typically consider already outdated models or only specific scenario(s). Consequently, there is a lack of a systematic study that benchmarks state-of-the-art models for a comprehensive set of programming education scenarios. In our work, we systematically evaluate two models, ChatGPT (based on GPT-3.5) and GPT-4, and compare their performance with human tutors for a variety of scenarios. We evaluate using five introductory Python programming problems and real-world buggy programs from an online platform, and assess performance using expert-based annotations. Our results show that GPT-4 drastically outperforms ChatGPT (based on GPT-3.5) and comes close to human tutors' performance for several scenarios. These results also highlight settings where GPT-4 still struggles, providing exciting future directions on developing techniques to improve the performance of these models.