Hyppää sisältöön
    • Suomeksi
    • På svenska
    • In English
  • Suomeksi
  • In English
  • Kirjaudu
Näytä aineisto 
  •   Etusivu
  • LUTPub
  • Diplomityöt ja Pro gradu -tutkielmat
  • Näytä aineisto
  •   Etusivu
  • LUTPub
  • Diplomityöt ja Pro gradu -tutkielmat
  • Näytä aineisto
JavaScript is disabled for your browser. Some features of this site may not work without it.

Prompting strategies and human-AI collaboration in a programming task

Himel, Aqemur Reza (2026)

Katso/Avaa
Mastersthesis_Himel_Aqemur_Reza.pdf (775.3Kb)
Lataukset: 


Diplomityö

Himel, Aqemur Reza
2026

School of Engineering Science, Tietotekniikka

Kaikki oikeudet pidätetään.
Näytä kaikki kuvailutiedot
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe20260716111376

Tiivistelmä

Generative AI coding assistants such as ChatGPT and GitHub Copilot are increasingly used for code generation, debugging and other programming tasks but it remains unclear how different prompting strategies affect human–AI collaboration and task outcomes. This thesis examines whether explicitly articulating one’s own understanding of a Python programming problem to an AI coding tool leads to different results than directly using the task description as the basis of the first prompt. The study adopts a between subjects experimental design with 147 participants with prior Python experience recruited through Prolific. Participants were randomly assigned to one of two conditions: direct prompting based on the task description or shared understanding oriented prompting in which they first described the problem in their own words. All participants completed the same authentication related Python task in a web based environment while using the given AI coding tool after which they responded to a post task questionnaire.

The study collected both objective and subjective data including the number of hidden test cases passed, binary task success, self reported time on task, number of prompts, acceptance of the first AI response, the self reported percentage of AI generated code in the final solution and questionnaire based measures of shared understanding, communication effort and perceived AI collaboration effectiveness. The experimental manipulation was successful. Participants in the direct prompting condition predominantly copied the task description into their first prompt whereas participants in the shared understanding condition consistently reformulated the task in their own words. However, the two conditions did not differ significantly in the number of test cases passed or in the probability of reaching the predefined binary success threshold. Instead, the shared understanding condition was associated with longer completion time and higher perceived communication effort while the direct prompting condition was associated with higher perceived AI collaboration effectiveness.

Correlation and regression analyses showed that the subjective constructs were meaningfully related to one another but only weakly related to objective performance. Shared understanding was positively associated with perceived AI collaboration effectiveness and negatively associated with communication effort yet neither condition nor the subjective measures emerged as strong multivariable predictors of task success. Overall, the findings suggest that requiring developers to explicitly explain their understanding to the AI changes how effortful the interaction feels and how long it takes but does not necessarily improve solution quality in a short, constrained programming task. The thesis contributes to research on AI assisted software development by experimentally isolating a concrete aspect of prompt design and by highlighting the distinction between perceived collaboration quality and objective coding outcomes.
Kokoelmat
  • Diplomityöt ja Pro gradu -tutkielmat [15768]
LUT-yliopisto
PL 20
53851 Lappeenranta
Ota yhteyttä | Tietosuoja | Saavutettavuusseloste
 

 

Tämä kokoelma

JulkaisuajatTekijätNimekkeetKoulutusohjelmaAvainsanatSyöttöajatYhteisöt ja kokoelmat

Omat tiedot

Kirjaudu sisäänRekisteröidy
LUT-yliopisto
PL 20
53851 Lappeenranta
Ota yhteyttä | Tietosuoja | Saavutettavuusseloste