Norman Peitek
Portrait of Norman Peitek

Dr. Norman Peitek

Postdoctoral Researcher, Chair of Software Engineering
Saarland University · Saarland Informatics Campus

About

AI assistants now write a substantial share of new code, yet developers remain accountable for the software they ship. They must understand the code that matters. My research bridges human and machine understanding of software. It is built around one core measure: the mental effort a developer needs to reason about code. I measure this effort in the brains of developers and compare it with how large language models process the same code.

I am a postdoctoral researcher in Sven Apel's ERC Advanced Grant Brains on Code at Saarland University.

During my doctorate at Chemnitz University of Technology and the Leibniz Institute for Neurobiology (summa cum laude, university dissertation award, nominated for the GI Dissertation Award), I developed a multimodal framework that measures program comprehension with fMRI, EEG, eye tracking, and psychophysiology. Our fMRI study on code complexity received an ACM SIGSOFT Distinguished Paper Award at ICSE, and a neurosurgical team recently drew on this work to preserve a software engineer's programming abilities during brain tumor surgery.

Before research, I worked as a software consultant on Android, Node.js, and .NET projects. As co-founder of Future Studio, I have written 200+ programming tutorials and 70 videos with more than 1.5 million views.

37 publications · 1'000+ citations · h-index 13 (Google Scholar, October 2026)

Research: Human and Machine Understanding of Software

My long-term goal is a linking of human reasoning processes and mechanistic representations in LLMs. Human data then become a reference signal to evaluate and improve AI assistants, and LLMs become testable theories of human program comprehension.

Area 1

Human understanding

Measuring and predicting program comprehension with fMRI, EEG, and eye tracking. We showed that code complexity metrics capture programmers' cognitive load only to a limited degree (ICSE 2021), that confusing code elicits a neural response resembling that to unexpected words (Sci. Rep. 2026), and that EEG separates high- from low-performing programmers (ESEC/FSE 2022).

Area 2

Machine understanding

Cognitively informed interpretability of code models. With probing, activation patching, and steering, we study how LLMs represent what makes code hard for developers: for example, where misleading names and comments derail LLMs (preprint 2026) and how LLMs converge when refactoring code for readability (preprint 2026).

Area 3

Bridging human and machine understanding

Comparing human and model processing of the same code. We showed that code that confuses programmers also tends to surprise LLMs, and that LLM-internal measures can point to code regions that are difficult for humans (ICSE 2026). Where humans and models diverge points to where generated code is likely to be misunderstood in review.

Selected Publications

  1. Are Humans and LLMs Confused by the Same Code? An Empirical Study on Fixation-Related Potentials and LLM Perplexity Youssef Abdelsalam†, Norman Peitek, Anna-Maria Maurer†, Mariya Toneva, and Sven Apel Proc. International Conference on Software Engineering (ICSE), pp. 2609–2621, ACM, 2026 ICSE 2026 PDF DOI
  2. Fixation-Related Potentials Reveal that Confusing Program Code Elicits a Late Frontal Positivity Annabelle Bergum†, Anna-Maria Maurer†, Norman Peitek, Regine Bader, Axel Mecklinger, Vera Demberg, Janet Siegmund, and Sven Apel Scientific Reports, 16:16833, 2026 Sci. Rep. 2026 PDF DOI
  3. On the Influence of the Baseline in Neuroimaging Experiments on Program Comprehension Annabelle Bergum†, Norman Peitek, Maurice Rekrut, Janet Siegmund, and Sven Apel ACM Transactions on Software Engineering and Methodology (TOSEM), 35(3):83, 2026 TOSEM 2026 PDF DOI
  4. Correlates of Programmer Efficacy and Their Link to Experience: A Combined EEG and Eye-Tracking Study Norman Peitek, Annabelle Bergum†, Maurice Rekrut, Jonas Mucke, Matthias Nadig, Chris Parnin, Janet Siegmund, and Sven Apel Proc. Joint Meeting on Foundations of Software Engineering (ESEC/FSE), pp. 120–131, ACM, 2022 ESEC/FSE 2022 PDF DOI
  5. Program Comprehension and Code Complexity Metrics: An fMRI Study Norman Peitek, Sven Apel, Chris Parnin, André Brechmann, and Janet Siegmund Proc. International Conference on Software Engineering (ICSE), pp. 524–536, IEEE, 2021 ICSE 2021 ACM SIGSOFT Distinguished Paper Award PDF DOI
  6. A Look into Programmers' Heads Norman Peitek, Janet Siegmund, Sven Apel, Christian Kästner, Chris Parnin, Anja Bethmann, Thomas Leich, Gunter Saake, and André Brechmann IEEE Transactions on Software Engineering (TSE), 46(4):442–462, 2020 TSE 2020 PDF DOI

† co-advised doctoral student · ‡ supervised thesis student

All 37 publications →

News

Awards

Teaching and Supervision

My teaching focuses on empirical thinking: judging code, tools, and claims by evidence rather than hype. Students design and run their own studies, work in teams with professional practices, and test claims about AI assistants against data.

Courses at Saarland University

Service

Organization

  • since 2022Steering Committee, International Workshop on Eye Movements in Programming (EMIP)
  • 2019, 2021, 2023–2026Organizer, EMIP
  • 2023Proceedings Co-Chair, ICPC

Reviewer

  • TSE, TOSEM, EMSE, TOCE, CHI, JSS, Human Brain Mapping, Cognitive Science, PLOS ONE, and others

Program committees

  • since 2024ICSE New Ideas and Emerging Results (NIER)
  • since 2024International Computing Education Research (ICER)
  • since 2024MSR Registered Reports
  • since 2023Human-Centric Experiments in Software Engineering (HUMAN)
  • 2018–2024ESEM, HFSD, EMIP, ETRA-ET4S, ICPC

Invited Talks