Skip to main navigation Skip to search Skip to main content

Evaluating One-Shot and Multi-Shot Prompting Strategies in a Transparent Educational Chatbot

  • Georgia Southern University
  • Dept. of Information Technology

Research output: Contribution to book or proceedingConference articlepeer-review

Abstract

This study explores how one-shot and multi-shot prompting strategies affect the performance of an educational chatbot powered by DeepSeek-R1, with a focus on reasoning transparency. The chatbot was evaluated using 30 educational questions across three categories: direct, comparison, and general knowledge, delivered through a Flask-based web application integrated with DeepSeek-R1 on Microsoft Azure. Multi-shot prompting used seven examples to provide richer context, while one-shot prompting relied on a single example. The evaluation employed automated metrics, including BERT-based semantic similarity, sentiment scores, and response length and timing. Results show that multi-shot prompting enhances clarity and conciseness for technical questions, while one-shot prompting performs better for open-ended queries. These findings suggest that aligning prompting strategies with question types can improve chatbot performance, support transparent reasoning, and enhance educational interactions.

Original languageEnglish
Title of host publicationProceedings - 29th IEEE/ACIS International Conference on Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing, SNPD 2025-Summer
EditorsHyun Yoe, Ha Jin Hwang, Meonghun Lee, Rackwoo Kim, Ryugap Lim, Sungtaek Lee, Seaeul Kim, Simon Xu, Miguel Garcia-Ruiz, Wenyin Feng, A B M Bodrul Alam, Randy Lin, Ajmery Sultana, Faria Khandaker, Mahreen Nasir, Ken Higuchi, Shinichiro Mori, Teruhisa Hochin
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages977-983
Number of pages7
ISBN (Electronic)9798331512583
DOIs
StatePublished - 2025
Event29th IEEE/ACIS International Conference on Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing, SNPD 2025-Summer - Busa, Korea, Republic of
Duration: Jun 25 2025Jun 27 2025

Publication series

NameProceedings - 29th IEEE/ACIS International Conference on Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing, SNPD 2025-Summer

Conference

Conference29th IEEE/ACIS International Conference on Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing, SNPD 2025-Summer
Country/TerritoryKorea, Republic of
CityBusa
Period06/25/2506/27/25

Scopus Subject Areas

  • Hardware and Architecture
  • Software
  • Information Systems and Management
  • Computer Vision and Pattern Recognition
  • Artificial Intelligence
  • Computer Networks and Communications

Keywords

  • AI
  • Chatbot
  • Large Language Models (LLMs)
  • Natural Language Processing (NLP)
  • Prompt Engineering

Fingerprint

Dive into the research topics of 'Evaluating One-Shot and Multi-Shot Prompting Strategies in a Transparent Educational Chatbot'. Together they form a unique fingerprint.

Cite this