From Passive Retrieval to Active Engagement: Designing and Evaluating Artificial Coding Buddy for Programming Education
Author
Koch, Daniela
Term
4. term
Education
Publication year
2026
Abstract
This thesis examines how general-purpose AI tools in programming education can lead to overreliance and a weakening of higher-order thinking skills that are essential for becoming an independent developer. In response, it presents Artificial Coding Buddy, an early prototype of a web-based AI chat system developed within the BuddhAI initiative at Aalborg University. Grounded in constructivist learning theory and Problem Based Learning, the prototype is designed to promote active and critical engagement with programming concepts rather than passive acceptance of generated code. The system offers three interaction modes – Explain, Generate Code, and Review Code – which respectively support conceptual understanding, require students to evaluate multiple code solutions of varying quality, and encourage reflection on their own code through articulated strengths, weaknesses, and related best practices. Each response is accompanied by follow-up questions to the user, targeting either lower- or higher-order thinking skills depending on their familiarity with the topic, as well as transparent access to the model’s underlying reasoning. The prototype was developed through an iterative process informed by student feedback across four testing cycles and then summatively evaluated by eight university lecturers experienced in Problem Based Learning. All experts acknowledged its educational value, and most considered it a preferable alternative to general-purpose AI tools for teaching; the Review Code mode, built-in scaffolding, and elements of gamification were highlighted as particular strengths. At the same time, key challenges were identified, including the risk that students may revert to more convenient general AI tools when faced with cognitive demands, the need for improved adaptability of question-based interactions, and usability issues that must be addressed before wider classroom deployment. The thesis outlines planned future work, including dedicated usability studies and a full-semester classroom trial, to inform the ongoing development of BuddhAI.
Denne afhandling undersøger, hvordan generelle AI-værktøjer i programmeringsuddannelse kan føre til overafhængighed og svækkelse af højere ordens tænkning, som er afgørende for at blive selvstændig udvikler. Som svar herpå præsenteres Artificial Coding Buddy, en tidlig prototype af et webbaseret AI-chatværktøj udviklet under BuddhAI-initiativet på Aalborg Universitet. Prototypen er forankret i konstruktivistisk læringsteori og Problem Baseret Læring og er designet til at fremme aktiv og kritisk engagement med programmeringsbegreber frem for passiv kopiering af genereret kode. Systemet tilbyder tre interaktionsmåder – Explain, Generate Code og Review Code – der henholdsvis støtter konceptuel forståelse, kræver vurdering af flere kodeforslag af varierende kvalitet, og faciliterer refleksion over egen kode via styrker, svagheder og relevante bedste praksisser. Hver svarsekvens ledsages af opfølgende spørgsmål til brugeren, målrettet enten lavere eller højere ordens tænkning afhængigt af deres forkundskaber, samt gennemsigtig visning af modellens ræsonnement. Prototypen blev udviklet iterativt med studenterfeedback over fire testrunder og efterfølgende summativt evalueret af otte universitetsundervisere med erfaring i Problem Baseret Læring. Alle eksperter vurderede værktøjet som pædagogisk værdifuldt, og de fleste foretrak det frem for generelle AI-chatbots til undervisningsbrug; særligt Review Code-tilstanden, den indbyggede stilladsering og gamifikationsaspekter blev fremhævet som styrker. Samtidig blev der identificeret udfordringer, herunder risikoen for at studerende vælger mere bekvemme generelle AI-værktøjer, behov for bedre tilpasning af de spørgebaserede interaktioner og usability-problemer, der skal adresseres før bred implementering. Afhandlingen peger på fremtidige studier af brugervenlighed og langvarig anvendelse i undervisning som næste skridt i videreudviklingen af BuddhAI.
[This abstract has been generated with the help of AI directly from the project full text]