Digitala Vetenskapliga Arkivet

Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Designing Interactive Machine Learning interfaces for non-expert users
Stockholm University, Faculty of Social Sciences, Department of Computer and Systems Sciences.
Stockholm University, Faculty of Social Sciences, Department of Computer and Systems Sciences.
2024 (English)Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesis
Abstract [en]

The development of artificial intelligence models is today in the hands of machine learning engineers and non-expert users rarely have an understanding of how the ‘black box’ works; they can only observe the input and output. This is problematic as only a limited group of individuals understands how to train an artificial intelligence model that may have the power to make and affect important decisions. The machine learning approach, interactive machine learning, may be a solution to this as it can allow for non-expert users to be a part of the process to train artificial intelligence models through an interactive machine learning interface. To allow a non-expert user to use such an interface, it must be usable and easy to understand which highlights the need of human-computer interaction in the design. A paradigm that has grown popular for analysis with artificial intelligence is satellite images, which has been used to detect both changes to land boundaries and natural disasters. This thesis explores how to design an interactive machine learning interface for non-expert users, enabling them to train artificial intelligence models according to their specific needs. The research question is therefore similarly How can a human-centered feedback interface for fine-tuning AI models be designed for non-expert users within the image analysis domain of satellite images? The goal of the thesis is to evaluate and test a graphical user interface which allows for non-expert users to annotate and correct data in satellite images to fine-tuning artificial intelligence models. The thesis was conducted with a design science approach as well as a survey approach with qualitative thematic analysis and three iterations of testing. First iteration involved evaluating a paper prototype with a domain expert. The second iteration included evaluating a Figma prototype through observation tests with non-experts. Finally, the third iteration involved testing of a new Figma prototype using a questionnaire, also with non-experts. The gathered data was transcribed and analyzed with an inductive approach which resulted in several themes, categories and codes. Thematic analysis revealed key aspects such as a high level of user confusion regarding interaction with the results of artificial intelligence models and a lack of understanding of certain icons' functionality. This highlights a greater need for guidance and the importance of following specific design principles when creating interactive machine learning interfaces. The conclusion emphasizes the necessity of integrating design principles and gestalt laws into interface design to facilitate effective communication and user guidance. By clearly communicating the task's goal through guiding and priming design, users can more effectively perform tasks. The proposed research question was partly answered, indicating that further evaluation of the prototype is needed to fully address the research question.

Place, publisher, year, edition, pages
2024.
Keywords [en]
Interactive Machine Learning, Human-computer interaction, Explainable AI, Non-expert, Human-centered design, Supervised learning
National Category
Computer Sciences
Identifiers
URN: urn:nbn:se:su:diva-242678OAI: oai:DiVA.org:su-242678DiVA, id: diva2:1955569
Available from: 2025-04-30 Created: 2025-04-30

Open Access in DiVA

fulltext(55770 kB)12 downloads
File information
File name FULLTEXT01.pdfFile size 55770 kBChecksum SHA-512
b7671fc4be14d2894711490136284354dada56b6c55e565db4dd8d505b9fb3f405cc3fcb56d15f1edd7cb22295cfdac914facb6d1c48514bf2164e3d4ae05a5e
Type fulltextMimetype application/pdf

Search in DiVA

By author/editor
Lundkvist, SofiaLinder Alexandersson, Yrsa
By organisation
Department of Computer and Systems Sciences
Computer Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 12 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 23 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf