HICSS 2023 : Hawaii International Conference on System Science: Minitrack AEO of AI - Accountability, Evaluation and Obscurity of AI Algorithms

We would like to invite papers, which would help us to understand how far we are from designing new computational models, which would evaluate past, current and future algorithms, which currently shape AI. In academia and industry, extensive discussions on biased data sets and algorithms triggered talks on achieving fairness, non-biased, risk-averse and trustworthy AI. We would be interested in learning how to synchronise technology advances, principles and practices of Data Science, abundance of data processed through AI, and automatic generation of algorithmic predictions through various software tools. Whether we are waiting for a new way of looking at the problem of transparency of AI or creating a new computational paradigm, which would automatically detect hidden dangers in data sets and associated algorithms, we need to engage in wider talks across communities which have impact on or are affected by the current status of AI.
We invite a spectrum of papers, which highlight any of our concerns and the topics below, and thus help us to cut through the current hype of AI. Discussions at the HICSS conference would answer many AI questions we may have and show a vision where the world of algorithmic computing in AI might be heading. We are particularly interested in the impact the Data Science. At the time of writing this proposal, we could not find a conference, which comprehensively addresses the way we create algorithmic computations, and thus would like to debate if Data Science practices might be contributing towards creating more obscurity of AI algorithms. The co-relation between Data Science practices and AI algorithms is obvious, but we still look at the two of them separately without evaluating the impact they may have on each other.
We would also welcome the applications of these novel research results in cyber physical spaces, pervasive and intelligent healthcare, medicine, robotics, smart cities and autonomous transport, automation, IoT and many more.
Indicative Topics are:
Evaluation of AI Algorithms
• Multidisciplinary nature of AI evaluations
• Methods of evaluating and explaining AI algorithms and their complexities
• Computational paradigm and frameworks for evaluating ML and AI Algorithms
• Reconstructing AI algorithms from their descriptions and explanations
• Automatic creation of transparent AI models and algorithms
• Software testing and AI: Testing AI and its algorithms (principles and methods)
• Adaptation of software testing to specificity of AI-based software solutions
• Interpretability of algorithms through models and criteria for decision making

The pace of AI innovation and responses of society
• Social and legal outcome of fair algorithms
• Intentioanl and unintentational consequences of using AI
• potential harm to people resulting from algorithmic opacity
• Human centred AI
• Human AI interaction: achieving fair, responsible, acceptable outcome of using AI
• AI for understanding humans and vice versa
• Ethics, legal and socio-technical issues in AI based systems
• Ethical issues and sharing principles in designing AI algorithms
• AI monitoring and regulations
• AI Algorithmic harm: dishonesty, fairness and lack of awareness in AI
• The risk of NOT using AI

Obscurity of AI Algorithms
• Impenetrable obscurity of AI
• AI evaluation and seeing “what is inside the black box”
• Black-box functions in AI and decision making
• Human control in AI versus black box algorithms versus transparency in AI
• Trade-off between algorithmic decision making in and transparency of AI
• AI accountability regime and the level of human trusts built by algorithms
• Accountability versus responsibility in AI based systems
• Performance of transparent AI models and algorithms

AI Algorithms, Fairness and Data Science
• Identifying BIAS which matters
• BIASED AI and automatic decision making
• BIASED semantic representation of data and human BIAS
• Abstractions in machine learning algorithms and fairness
• Impact of Data Science on semantic of data in training/learning algorithms
• Feature selection/engineering in ML versus fairness of algorithms where it is used
• Impact of feature selection on the testing/ training data sets which may affects fairness and discover harms of AI
• Co-relation between Data Science practices and AI computational models
• Discovering fairness in AI algorithms through the semantic stored in data sets
• Impact of statistical methods on Data Science and on the evaluation/transparency/ obscurity of AI

Co-Chairs

Dr Radmila Juric, PhD ([email protected] )
ALMAIS Consultancy, London, UK, [email protected]

Dr Robert J. Steele, Ph.D.
Chair and Full Professor of Computer Science
Department of Computer Science, Capitol Technology University
11301 Springfield Rd, Laurel, MD 20708, USA, [email protected]