Three Levels of AI Transparency

Kashyap Haresamudram; Stefan Larsson; Fredrik Heintz

doi:10.1109/MC.2022.3213181

Three Levels of AI Transparency

Kashyap Haresamudram, Stefan Larsson, Fredrik Heintz

Linköping University

Research output: Contribution to journal › Article › peer-review

2200 Downloads (Pure)

Abstract

Transparency is generally cited as a key consideration towards building Trustworthy AI. However, the concept of transparency is fragmented in AI research, often limited to transparency of the algorithm alone. While considerable attempts have been made to expand the scope beyond the algorithm, there has yet to be a holistic approach that includes not only the AI system, but also the user, and society at large. We propose that AI transparency operates on three levels, (1) Algorithmic Transparency, (2) Interaction Transparency, and (3) Social Transparency, all of which need to be considered to build trust in AI. We expand upon these levels using current research directions, and identify research gaps resulting from the conceptual fragmentation of AI transparency highlighted within the context of the three levels.

Original language	English
Pages (from-to)	93-100
Journal	Computer
Volume	56
Issue number	2
DOIs	https://doi.org/10.1109/MC.2022.3213181
Publication status	Published - 2023

Subject classification (UKÄ)

Other Engineering and Technologies
Other Social Sciences
Information Systems, Social aspects (including Human Aspects of ICT)

Free keywords

Artificial Intelligence
Transparency
Algorithm
Interaction
Society
Governance

Access to Document

10.1109/MC.2022.3213181

Three Levels of AI TransparencyAccepted author manuscript, 562 KB

1 Book chapter
1 Doctoral Thesis (compilation)

Interactions with Pseudo-Sapiens: User perception of anthropomorphism, mind, and trust in humanlike social agents
Haresamudram, K., 2025 Jan 20, Lund, Sweden: Department of Technology and Society, Lund University. 112 p.
Research output: Thesis › Doctoral Thesis (compilation)

Open Access
File
Consumer Trust and Platformised Retail Personalisation
Larsson, S. & Haresamudram, K., 2023 Oct 9, The Future of Consumption: How Technology, Sustainability and Wellbeing will Transform Retail and Customer Experience. Bäckström, K., Egan-Wyer, C. & Samsioe, E. (eds.). Palgrave Macmillan, p. 77-94 18 p.
Research output: Chapter in Book/Report/Conference proceeding › Book chapter › Research › peer-review

Open Access
File
30 Downloads (Pure)

2 Finished

Automated decision-making – Nordic perspectives
Larsson, S. (Researcher), Ledendal, J. (Researcher), de Vries, K. (Researcher), Heintz, F. (Researcher), Ruckenstein, M. (Researcher), Lomborg, S. (PI), Henriksen, A. (Researcher), Söderlund, K. (Research student) & Högberg, C. (Research student)
2020/02/03 → 2022/09/30
Project: Research
AI Transparency and Consumer Trust
Larsson, S. (PI), Heintz, F. (Researcher), Söderlund, K. (Research student), Haresamudram, K. (Research student), Tanqueray, L. (Project coordinator) & White, J. (Researcher)
2020/01/01 → 2024/12/31
Project: Research

4 Invited talk

The Role of Transparency in the Governance of Artificial Intelligence
Larsson, S. (Keynote/plenary speaker)
2024 Jun 12
Activity: Talk or presentation › Invited talk
Normer och socialt hållbar AI
Larsson, S. (Invited speaker)
2023 May 9
Activity: Talk or presentation › Invited talk
AI and Norms ....and interdisciplinary struggles
Larsson, S. (Keynote/plenary speaker)
2023 Apr 27
Activity: Talk or presentation › Invited talk

Cite this

@article{9405ca170949443e8cad977f38c2df4f,

title = "Three Levels of AI Transparency",

abstract = "Transparency is generally cited as a key consideration towards building Trustworthy AI. However, the concept of transparency is fragmented in AI research, often limited to transparency of the algorithm alone. While considerable attempts have been made to expand the scope beyond the algorithm, there has yet to be a holistic approach that includes not only the AI system, but also the user, and society at large. We propose that AI transparency operates on three levels, (1) Algorithmic Transparency, (2) Interaction Transparency, and (3) Social Transparency, all of which need to be considered to build trust in AI. We expand upon these levels using current research directions, and identify research gaps resulting from the conceptual fragmentation of AI transparency highlighted within the context of the three levels.",

keywords = "Artificial Intelligence, Transparency, Algorithm, Interaction, Society, Governance",

author = "Kashyap Haresamudram and Stefan Larsson and Fredrik Heintz",

year = "2023",

doi = "10.1109/MC.2022.3213181",

language = "English",

volume = "56",

pages = "93--100",

journal = "Computer",

issn = "1558-0814",

publisher = "IEEE Computer Society",

number = "2",

}

TY - JOUR

T1 - Three Levels of AI Transparency

AU - Haresamudram, Kashyap

AU - Larsson, Stefan

AU - Heintz, Fredrik

PY - 2023

Y1 - 2023

N2 - Transparency is generally cited as a key consideration towards building Trustworthy AI. However, the concept of transparency is fragmented in AI research, often limited to transparency of the algorithm alone. While considerable attempts have been made to expand the scope beyond the algorithm, there has yet to be a holistic approach that includes not only the AI system, but also the user, and society at large. We propose that AI transparency operates on three levels, (1) Algorithmic Transparency, (2) Interaction Transparency, and (3) Social Transparency, all of which need to be considered to build trust in AI. We expand upon these levels using current research directions, and identify research gaps resulting from the conceptual fragmentation of AI transparency highlighted within the context of the three levels.

AB - Transparency is generally cited as a key consideration towards building Trustworthy AI. However, the concept of transparency is fragmented in AI research, often limited to transparency of the algorithm alone. While considerable attempts have been made to expand the scope beyond the algorithm, there has yet to be a holistic approach that includes not only the AI system, but also the user, and society at large. We propose that AI transparency operates on three levels, (1) Algorithmic Transparency, (2) Interaction Transparency, and (3) Social Transparency, all of which need to be considered to build trust in AI. We expand upon these levels using current research directions, and identify research gaps resulting from the conceptual fragmentation of AI transparency highlighted within the context of the three levels.

KW - Artificial Intelligence

KW - Transparency

KW - Algorithm

KW - Interaction

KW - Society

KW - Governance

U2 - 10.1109/MC.2022.3213181

DO - 10.1109/MC.2022.3213181

M3 - Article

SN - 1558-0814

VL - 56

SP - 93

EP - 100

JO - Computer

JF - Computer

IS - 2

ER -

Three Levels of AI Transparency

Abstract

Subject classification (UKÄ)

Free keywords

Access to Document

Fingerprint

Research output

Interactions with Pseudo-Sapiens: User perception of anthropomorphism, mind, and trust in humanlike social agents

Consumer Trust and Platformised Retail Personalisation

Projects

Automated decision-making – Nordic perspectives

AI Transparency and Consumer Trust

Activities

The Role of Transparency in the Governance of Artificial Intelligence

Normer och socialt hållbar AI

AI and Norms ....and interdisciplinary struggles

Cite this