Analisis Workarounds Generative AI dalam Menyusun Tugas Akhir

Buana, Akbar Daniswara Cahya (2026) Analisis Workarounds Generative AI dalam Menyusun Tugas Akhir. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026221202-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Perkembangan Generative Artificial Intelligence (GenAI) telah mengubah cara mahasiswa tingkat akhir bidang pemrograman menyusun tugas akhir, bukan sekadar sebagai alat bantu teknis, melainkan sebagai mitra interaktif yang menuntut strategi adaptasi aktif ketika keluarannya tidak memenuhi kebutuhan teknis maupun akademik. Penelitian ini bertujuan mendeskripsikan dan menganalisis bentuk, karakteristik, serta tahapan workarounds yang diterapkan mahasiswa ketika menggunakan GenAI dalam proses penyusunan tugas akhir pemrograman. Pendekatan studi kasus kualitatif deskriptif mengacu pada Eisenhardt (1989) diterapkan terhadap 30 mahasiswa tingkat akhir dari lima perguruan tinggi di Indonesia. Data dikumpulkan melalui dua layer yang dianalisis secara terpisah, yaitu transkrip wawancara semi-terstruktur yang menghasilkan 502 segmen terkodekan dan dokumentasi riwayat prompt-response yang menghasilkan 511 segmen terkodekan, dengan total 1.013 unit analisis. Proses kodifikasi tiga level menghasilkan 52 Third Order codes yang dikelompokkan ke dalam tujuh pola utama. Temuan penelitian mengidentifikasi model dua fase penggunaan GenAI: Fase Orientasi yang mencakup penetapan arah pengerjaan melalui AI, manajemen konteks dan sesi, konstruksi instruksi awal, serta konfigurasi perilaku AI; dan Fase Siklus Workaround yang mencakup penyempurnaan prompt secara iteratif, verifikasi output, dan respons terhadap keterbatasan AI. Penyempurnaan Prompt Secara Iteratif merupakan satu-satunya perilaku yang hadir secara universal pada seluruh 30 partisipan di kedua layer data. Perbandingan lintas layer mengungkap kesenjangan sistematis antara perilaku yang dinarasikan dalam wawancara dan yang terekam dalam riwayat prompt, dengan kesenjangan terbesar pada pola respons keterbatasan AI sebesar 17 partisipan dan pola verifikasi output sebesar 12 partisipan. Konfigurasi perilaku AI teridentifikasi sebagai dimensi pembeda terkuat antar partisipan. Dari temuan ini dirumuskan dua proposisi konseptual yang menjelaskan mekanisme pembentukan workaround serta pengaruh kualitas orientasi awal terhadap karakter siklus workaround.
========================================================================================================================================
The advancement of Generative Artificial Intelligence (GenAI) has reshaped how final-year programming students complete their undergraduate thesis, positioning this technology not merely as a technical tool but as an interactive partner that demands active adaptation strategies when its outputs fail to meet technical or academic needs. This study aims to describe and analyze the forms, characteristics, and workflow stages of workarounds applied by students when using GenAI during the preparation of their programming undergraduate thesis. A qualitative descriptive case study approach following Eisenhardt (1989) was employed with 30 final-year students from five universities in Indonesia. Data were collected through two separately analyzed layers: semi-structured interview transcripts yielding 502 coded segments and prompt-response history documentation yielding 511 coded segments, totaling 1,013 units of analysis. A three-level coding process produced 52 Third Order codes grouped into seven main patterns. The findings identify a two-phase model of GenAI use: an Orientation Phase encompassing task direction-setting through AI, context and session management, initial instruction construction, and AI behavior configuration; and a Workaround Cycle Phase encompassing iterative prompt refinement, output verification, and response to AI limitations. Iterative Structured Prompting was the only universal behavior present across all 30 participants in both data layers. Cross-layer comparison revealed systematic gaps between behaviors narrated in interviews and those documented in prompt histories, with the largest gaps in the AI limitation response pattern at 17 participants and the output verification pattern at 12 participants. AI behavior configuration was identified as the strongest differentiating dimension across participants. Two conceptual propositions were formulated to explain the mechanisms of workaround formation and the influence of orientation phase quality on workaround cycle character.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Generative Artificial Intelligence, Workarounds, Studi Kasus Kualitatif, Workflow Prompting, Tugas Akhir Pemrograman. Generative AI, Workarounds, Prompt Engineering, Workflow Prompting, Programming Thesis.
Subjects:	H Social Sciences > HD Industries. Land use. Labor > HD45 Technological innovations L Education > LB Theory and practice of education > LB1044.87 Internet in education (e-learning). Virtual reality in education. L Education > LB Theory and practice of education > LB2300 Higher Education
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Akbar Daniswara Cahya Buana
Date Deposited:	29 Jul 2026 02:38
Last Modified:	29 Jul 2026 02:38
URI:	http://repository.its.ac.id/id/eprint/138511

Actions (login required)

View Item