Ändra sökning
RefereraExporteraLänk till posten
Permanent länk

Direktlänk
Referera
Referensformat
  • apa
  • harvard1
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Annat format
Fler format
Språk
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Annat språk
Fler språk
Utmatningsformat
  • html
  • text
  • asciidoc
  • rtf
Creating Designs of Future Systems with Interpretation of Cognitive Artifacts in Reinforcement Learning
KTH, Skolan för kemi, bioteknologi och hälsa (CBH), Medicinteknik och hälsosystem, Hälsoinformatik och logistik.ORCID-id: 0000-0003-1126-3781
(Engelska)Ingår i: Design Science Journal, ISSN 2053-4701Artikel i tidskrift (Refereegranskat) Submitted
Abstract [en]

Designing future systems such as transport or healthcare in a city takes astute expertise. Design aids in such situations usually offer information in the form of projections or what-if analysis, using which experts make a series of decisions to create bounded designs. We present a case in which Reinforcement Learning (RL) is used to design the future transport system of a city. RL is used to create artifacts that reflect where the transport system can be changed. These agent-produced artifacts are then compared with designs made by human experts. This is achieved by analogizing the city as gridworld and using the same information that the human experts acted on as rewards. The interpretation of agent activity as cognitive artifacts of agents, along with measures of precision and recall to compare real and artificial artifacts form the basis of this work. This paper explores the use of RL in a real world context and the interpretability of results of RL with respect to design problems. The results indicate a robust initial approach to imitating expertise of designers and devising valid creativity in Socio-Technical Systems.

Nyckelord [en]
Design Science, Mimicking, Creativity, Interpretability, Socio-Technical Systems
Nationell ämneskategori
Elektroteknik och elektronik
Identifikatorer
URN: urn:nbn:se:kth:diva-259568OAI: oai:DiVA.org:kth-259568DiVA, id: diva2:1352182
Tillgänglig från: 2019-09-17 Skapad: 2019-09-17 Senast uppdaterad: 2019-09-20
Ingår i avhandling
1. Creating Knowledge with Data Science for Design in Systems
Öppna denna publikation i ny flik eller fönster >>Creating Knowledge with Data Science for Design in Systems
2019 (Engelska)Doktorsavhandling, sammanläggning (Övrigt vetenskapligt)
Abstract [en]

Designing in large-scale engineering systems is a difficult cognitive task undertaken by experts. Knowledge of experts continually changes as they are confronted with similar by different problems in designing in such systems. However, it is also important that designers are presented information that is representative of the system,and that they are cognizant of activities on a system scale so they can create diverse choices for designs in early phase of design process.Data Science has been proven to be effective at informing people for decisions at immediate horizons. But the use of data science to drive long terms designs where experts have to make the right series of decisions i.e. designs is yet unknown. The use of data science is to inform decision makers of activities at system scale.In this thesis, I have looked at how data science can be used to create knowledge in designers for designing in large scale systems. I have also investigated further questions regarding imitation of expertise using AI, and in generating similar knowledge by creating diverse options in design.The results point out that data science can indeed inform designers, change their designs and hence create knowledge. They also point out that design cognition in experts can be partly imitated in data science itself, through careful modeling of the ill-defined problem in design. This therefore points to a promising future direction where data can be used as an interface between human thinking and machine learning, by translation of conceptual forms such as differential diagnoses and cognitive artefacts using data.

Ort, förlag, år, upplaga, sidor
KTH Royal Institute of Technology, 2019. s. 33
Serie
TRITA-CBH-FOU ; 46
Nyckelord
Data Science, Design Cognition, Transport, Healthcare, Artificial Intelligence, Imitation, Design Space Exploration
Nationell ämneskategori
Datorsystem Hälsovetenskaper Transportteknik och logistik
Forskningsämne
Teknik och hälsa
Identifikatorer
urn:nbn:se:kth:diva-259566 (URN)978-91-7873-299-9 (ISBN)
Disputation
2019-10-11, T2, Hälsovägen 11, 141 57, Huddinge, 09:30 (Engelska)
Opponent
Handledare
Anmärkning

QC 2019-09-20

Tillgänglig från: 2019-09-20 Skapad: 2019-09-18 Senast uppdaterad: 2019-10-18Bibliografiskt granskad

Open Access i DiVA

Fulltext saknas i DiVA

Personposter BETA

Shreenath, Vinutha MagalMeijer, Sebastiaan

Sök vidare i DiVA

Av författaren/redaktören
Shreenath, Vinutha MagalMeijer, Sebastiaan
Av organisationen
Hälsoinformatik och logistik
Elektroteknik och elektronik

Sök vidare utanför DiVA

GoogleGoogle Scholar

urn-nbn

Altmetricpoäng

urn-nbn
Totalt: 20 träffar
RefereraExporteraLänk till posten
Permanent länk

Direktlänk
Referera
Referensformat
  • apa
  • harvard1
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Annat format
Fler format
Språk
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Annat språk
Fler språk
Utmatningsformat
  • html
  • text
  • asciidoc
  • rtf