AI Governance Through a Transparency Lens. Theodorou, A. Ph.D. Thesis, University of Bath, 2019. ![AI Governance Through a Transparency Lens [pdf]](https://bibbase.org/img/filetypes/pdf.svg "pdf")
Paper abstract bibtex Transparency is a key consideration for the ethical design and use of Artificial Intelligence, and has recently become a topic of considerable public interest and debate. We frequently use philosophical, mathematical, and biologically inspired techniques for building artificial, interactive, intelligent agents. Yet despite these well-motivated inspirations, the resulting intelligence is often developed as a black box, communicating no understanding of how the underlying real-time decision making functions. This compromises both the safety of such systems and fair attribution of moral responsibility and legal accountability when incidents occur. This dissertation provides the knowledge and software tools to make artificially intelligent agents more transparent, allowing a direct understanding of the action-selection system of such a system. The use of transparency, as demonstrated in this document, helps not only with the debugging of intelligent agents, but also with the public's understanding of Artificial Intelligence (AI) by removing the 'scary' mystery around "why is it behaving like that". In the research described in this document I investigate and compare the perception we have of intelligent systems, such as robots and autonomous vehicles, when they are treated as black boxes compared to when we make their action-selection systems transparent. Finally, I make normative and descriptive arguments for the moral status of intelligent systems and contribute to regulatory policy regarding such systems.
@phdthesis{Theodorou2019Thesis,
abstract = {Transparency is a key consideration for the ethical design and use of Artificial Intelligence, and has recently become a topic of considerable public interest and debate. We frequently use philosophical, mathematical, and biologically inspired techniques for building artificial, interactive, intelligent agents. Yet despite these well-motivated inspirations, the resulting intelligence is often developed as a black box, communicating no understanding of how the underlying real-time decision making functions. This compromises both the safety of such systems and fair attribution of moral responsibility and legal accountability when incidents occur. This dissertation provides the knowledge and software tools to make artificially intelligent agents more transparent, allowing a direct understanding of the action-selection system of such a system. The use of transparency, as demonstrated in this document, helps not only with the debugging of intelligent agents, but also with the public's understanding of Artificial Intelligence (AI) by removing the 'scary' mystery around "why is it behaving like that". In the research described in this document I investigate and compare the perception we have of intelligent systems, such as robots and autonomous vehicles, when they are treated as black boxes compared to when we make their action-selection systems transparent. Finally, I make normative and descriptive arguments for the moral status of intelligent systems and contribute to regulatory policy regarding such systems.},
author = {Theodorou, Andreas},
url = {{https://github.com/RecklessCoding/recklesscoding.github.io/raw/master/files/andreasTheodorouThesis.pdf}},
school = {University of Bath},
title = {{AI Governance Through a Transparency Lens}},
keywords = {User Studies, Systems AI, Cognitive Architectures, XAI, Transparency, Accountability, Governance, Policy, HRI, Human Control, Trust, Robotics},
year = {2019}
}
Downloads: 0
{"_id":"NwXps4qQ8x4Mq9dxc","bibbaseid":"theodorou-aigovernancethroughatransparencylens-2019","authorIDs":["5d29a84e67cf7dda01000060","5d3ee7bcf193c4da01000003","5de91b265d589edf01000062","5dea1156fac96fde01000195","5dee4372eaaee4df010000cb","5df0c15e8367c8de01000166","5df8a57910b1d1de010000e4","5dfe427168d95dde010000b0","5e1c852b4c9abfde0100006c","5e1cf2b4633f2ade01000038","5e232e1c327a15de010000c1","5e2424613abca2de01000069","5e2427323abca2de01000084","5e2f16684557f3de0100006f","5e305c7057a222df010001fb","5e3d2ad5c405ecde01000089","5e405a67aeea22df01000075","5e4077c9b531d7de01000095","5e439656639a35de010000fc","5e4668f7696958de010000ce","5e49462b379644de01000061","5e4ab1bd15f6c7df0100011b","5e4af24941072bdf01000161","5e4d7edd08a8e5de0100023b","5e505f7acb6c3ede01000043","5e51a9962793ecde01000115","5e5a3aea990a84ee010000bb","5e5dd5c9c64f0ede01000120","5e5fcac06b32b0f201000123","5e6a87e50e8744de0100007b","6GKzRsYg5ovkwuA4c","6jxJ88mC6PRaoFrYC","ADCDX8fp7q4kfMbiF","Augo264uoCi2XzYbq","EoyW3pNnENKsdXyQR","JJjuehbapGvrNARiB","LbTM6G6MWhf8NzcjG","MHYPbxh2Te5n46krF","MMrdK4f9fhov9aizv","Pmobwi7uoSEHwuZfW","RMFJdRjrLfqb27CMH","TqRoEcjhSeSNcgGzN","TrrPcb7aBYNoa9qAA","XPHjZdrT6b7SaALfi","XWvqr4MZaCCJkhexb","YZpuKuxccikQL9H4W","aRXLakuAEygxzR69H","baeKsXAuiXQA6kwL7","eYj3LbZx3NcTmPqEb","fhdGP7eL4naG5b2ju","oLmQdp9NYrXK9ecqC","qW4p25qJLKEHXwTkB","sPt7NArjdJJd5H4oA","tmLMcAELZL4q7BX2N","us4de8R6f9LCGNqmB","xoyhTiYvFPfAjJFka","yxC6FSkydyF4Z375C"],"author_short":["Theodorou, A."],"bibdata":{"bibtype":"phdthesis","type":"phdthesis","abstract":"Transparency is a key consideration for the ethical design and use of Artificial Intelligence, and has recently become a topic of considerable public interest and debate. We frequently use philosophical, mathematical, and biologically inspired techniques for building artificial, interactive, intelligent agents. Yet despite these well-motivated inspirations, the resulting intelligence is often developed as a black box, communicating no understanding of how the underlying real-time decision making functions. This compromises both the safety of such systems and fair attribution of moral responsibility and legal accountability when incidents occur. This dissertation provides the knowledge and software tools to make artificially intelligent agents more transparent, allowing a direct understanding of the action-selection system of such a system. The use of transparency, as demonstrated in this document, helps not only with the debugging of intelligent agents, but also with the public's understanding of Artificial Intelligence (AI) by removing the 'scary' mystery around \"why is it behaving like that\". In the research described in this document I investigate and compare the perception we have of intelligent systems, such as robots and autonomous vehicles, when they are treated as black boxes compared to when we make their action-selection systems transparent. Finally, I make normative and descriptive arguments for the moral status of intelligent systems and contribute to regulatory policy regarding such systems.","author":[{"propositions":[],"lastnames":["Theodorou"],"firstnames":["Andreas"],"suffixes":[]}],"url":"https://github.com/RecklessCoding/recklesscoding.github.io/raw/master/files/andreasTheodorouThesis.pdf","school":"University of Bath","title":"AI Governance Through a Transparency Lens","keywords":"User Studies, Systems AI, Cognitive Architectures, XAI, Transparency, Accountability, Governance, Policy, HRI, Human Control, Trust, Robotics","year":"2019","bibtex":"@phdthesis{Theodorou2019Thesis,\n\tabstract = {Transparency is a key consideration for the ethical design and use of Artificial Intelligence, and has recently become a topic of considerable public interest and debate. We frequently use philosophical, mathematical, and biologically inspired techniques for building artificial, interactive, intelligent agents. Yet despite these well-motivated inspirations, the resulting intelligence is often developed as a black box, communicating no understanding of how the underlying real-time decision making functions. This compromises both the safety of such systems and fair attribution of moral responsibility and legal accountability when incidents occur. This dissertation provides the knowledge and software tools to make artificially intelligent agents more transparent, allowing a direct understanding of the action-selection system of such a system. The use of transparency, as demonstrated in this document, helps not only with the debugging of intelligent agents, but also with the public's understanding of Artificial Intelligence (AI) by removing the 'scary' mystery around \"why is it behaving like that\". In the research described in this document I investigate and compare the perception we have of intelligent systems, such as robots and autonomous vehicles, when they are treated as black boxes compared to when we make their action-selection systems transparent. Finally, I make normative and descriptive arguments for the moral status of intelligent systems and contribute to regulatory policy regarding such systems.},\n\tauthor = {Theodorou, Andreas},\n\turl = {{https://github.com/RecklessCoding/recklesscoding.github.io/raw/master/files/andreasTheodorouThesis.pdf}},\n\tschool = {University of Bath},\n\ttitle = {{AI Governance Through a Transparency Lens}},\n\tkeywords = {User Studies, Systems AI, Cognitive Architectures, XAI, Transparency, Accountability, Governance, Policy, HRI, Human Control, Trust, Robotics},\n\tyear = {2019}\n}\n\n","author_short":["Theodorou, A."],"key":"Theodorou2019Thesis","id":"Theodorou2019Thesis","bibbaseid":"theodorou-aigovernancethroughatransparencylens-2019","role":"author","urls":{"Paper":"https://github.com/RecklessCoding/recklesscoding.github.io/raw/master/files/andreasTheodorouThesis.pdf"},"keyword":["User Studies","Systems AI","Cognitive Architectures","XAI","Transparency","Accountability","Governance","Policy","HRI","Human Control","Trust","Robotics"],"metadata":{"authorlinks":{}}},"bibtype":"phdthesis","creationDate":"2019-07-13T10:02:11.333Z","downloads":0,"keywords":["user studies","systems ai","cognitive architectures","xai","transparency","accountability","governance","policy","hri","human control","trust","robotics"],"search_terms":["governance","through","transparency","lens","theodorou"],"title":"AI Governance Through a Transparency Lens","year":2019,"biburl":"https://raw.githubusercontent.com/RecklessCoding/recklesscoding.github.io/master/files/MyPublications.bib","dataSources":["nakTuHcBPDNcWRCar"]}