Eindhoven
University of
Technology

Background information

In the Urban Systems and Real Estate (USRE) unit at TU/e, students often conduct survey-based research as part of their education. In courses like Smart Mobility Design (Bachelor) and the Graduation Project (Master), students gather insights into societal needs and behaviors to design user-centric solutions. However, collecting enough high-quality survey responses is often time-consuming and expensive. This challenge can lead to incomplete datasets, rushed conclusions, and reduced educational value. In many cases, the lack of timely data forces students to rely on theoretical assumptions rather than empirical analysis.

To address this issue, the LLM4survey project proposes the use of large language models (LLMs) such as ChatGPT and Gemini to generate representative survey responses. These AI-generated datasets are not intended to replace real data but to supplement and accelerate the research process. By integrating these tools, students can start analyzing data earlier in their project timelines and explore a wider range of scenarios, including ethically sensitive or hard-to-reach topics. The innovation helps reduce project delays, boosts student engagement, and increases the quality of data-driven learning experiences. It also aligns with TU/e’s broader goals of promoting flexibility, hybrid learning, and digital innovation.

Aim of the project

The goal of the LLM4survey project is to enhance survey-based student research by integrating LLMs as a supplementary source of representative data. This allows students to conduct deeper analyses without the delays typically associated with collecting sufficient human responses. The project focuses on improving the efficiency, flexibility, and quality of educational outcomes in two pilot courses: Smart Mobility Design and the USRE Graduation Project.

Students will be trained to use LLMs to generate representative datasets using structured prompts. These datasets will complement traditional survey data, making it possible to test hypotheses and analyze complex questions earlier in the project timeline. The tool also allows exploration of scenarios that are ethically sensitive or difficult to address through conventional survey methods. By incorporating AI in this way, students gain a more complete and motivating research experience while also developing skills in handling emerging technologies responsibly.

The project will evaluate the realism and validity of the AI-generated data and its impact on student learning, motivation, and project outcomes. It will also assess workload reduction for both students and supervisors. Ultimately, the project aims to develop reusable guidelines and best practices so the approach can be scaled across other courses and institutions that face similar challenges in data collection.

For more information, please contact:

Assistant Professor
Feixiong Liao
Built Environment
+31 40 2474037
University Lecturer
Peter van der Waerden
Built Environment