Generative artificial intelligence or Gen AI has transformed the experiences of both students and instructors in higher education. Researchers from the University of Southern California Center for Generative AI and Society provide a comprehensive examination of this phenomenon through the combination of surveys and experiments across several countries.
Executive Help or Instrumental Help: How Guidance and Scaffolding Can Shape Student AI Use
Background
The specific research team included Stephen J. Aguilar of the USC Rossier School of Education and William Swartout of USC Institute for Creative Technologies. Their work was motivated by concerns that many students were using generative AI primarily to avoid meaningful intellectual engagement despite its potential to enrich the learning experience.
Researchers pursued the study to identify patterns of student and teacher behavior, and to test whether generative AI tools could be designed to encourage reflection instead of shortcutting. They also aimed to inform institutional policies, instructor training, and tool development in order to support equitable and effective use of artificial intelligence in learning.
Methodology was multi-layered. One survey gathered responses from about 1000 college students in the United States to examine how generative AI was used in academic work. A second survey included 1505 teachers from the United States, India, Qatar, Colombia, and the Philippines. An experimental trial was also conducted to test an AI writing tool.
The student survey measured two categories of use. “Executive help” describes effort-saving tasks like producing summaries or answers with minimal thinking. “Instrumental help” describes deeper uses like clarifying difficult concepts, enhancing arguments, or developing skills. The survey also asked about professor encouragement or discouragement of AI.
Moreover, for the teachers, their beliefs or attitudes about benefits, risks, and institutional support were assessed. Respondents reported whether they thought artificial intelligence tools enhanced efficiency, personalized instruction, or learning outcomes, and whether they feared plagiarism, reduced creativity, or inconsistent guidance from universities.
The experimental component of the study evaluated a writing assistant AI for Brainstorming and Editing or ABE. This tool does not generate full drafts, unlike other systems. It was designed to encourage revision and reflection. Students were asked to use the tool in writing tasks, and researchers recorded patterns of interaction and perceived outcomes.
Key Findings
Dominant Executive Use: Most students used generative AI for executive help. This means that AI tools were used to finish tasks quickly or reduce workload, rather than to engage critically with assignments. This pattern suggests a strong preference for efficiency over educational depth when guidance is absent.
Limited Instrumental Use: Fewer students relied on AI for instrumental purposes like improving comprehension, clarifying concepts, or strengthening writing. Instrumental use represented the potential of AI to reinforce learning, but survey results showed that this opportunity was largely underutilized by the student population.
Professor Guidance Effect: Students whose professors actively encouraged responsible use of AI were significantly more likely to apply AI tools in instrumental ways. Instructor engagement and course design therefore played a major role in shifting student behavior from avoidance of learning toward deeper academic practices.
Wide Teacher Optimism: The international survey showed optimism. 72 percent said AI streamlines routine tasks. 73 percent said it improves outcomes. 69 percent noted its potential for personalized learning. Concerns about plagiarism, diminished creativity, and lack of institutional resources were high among educators.
ABE Tool as Companion: Students using ABE reported treating it as a companion to their writing, rather than a substitute. They used it to explore counterarguments, clarify claims, and expand perspectives. This suggested that properly designed tools could nudge students toward reflection, revision, and stronger critical engagement.
Implications
The study offers one of the most detailed looks at how generative artificial intelligence tools are transforming education worldwide. Moreover, by distinguishing between superficial executive help and deeper instrumental help, the researcher highlighted both the risks of relying on shortcuts and the potential for AI to enhance learning through guidance.
Its findings reveal that AI is not inherently detrimental to education. Its benefits or costs depend heavily on context and design. Left unguided, students often gravitate toward executive uses that sidestep intellectual growth. However, when professors offer direction, and when tools include scaffolding, AI becomes a facilitator for instrumental learning.
Educators face both opportunities and risks. There is a need to invest in professional development to equip instructors with strategies for effective AI integration. Without consistent guidance and resources, inequities may widen, and students may default to superficial uses. With thoughtful design, however, AI can genuinely expand learning possibilities.
FURTHER READING AND TAKEAWAY
- Aguilar, S. J., Nye, B., Swartout, W., Macias, A., Xing, Y., and Le Xiu, R. 2025. “How Students and Teachers Worldwide Are Adapting to AI. Center for Open Science.: EdArXiv Preprints. DOI: 35542/osf.io/wr6n3_v2