The ethics of artificially intelligent algorithms in society, public libraries and archives 

Our society is increasingly controlled by artificially intelligent, pervasive software systems that feed on our neverending growing heap of data traces that represent our characteristics, our wishes, our desires, our beliefs, and even our darkest secrets. Well-known examples are giant tech-firms like Google, Twitter, and Facebook, but there are numerous digital parts of our society where algorithms are being employed to monitor us, predict us, manipulate us, and to control us. Such systems increasingly replace interactions that were once purely physical. For example, Amazon's Kindle e-reader now intermediates aspects of buying, browsing, reading, and marking up, books. The dating app Tinder replaces aspects of finding, meeting and looking at potential partners. Facebook mediates many once-physical activities, such as talking with people, showing photographs, commenting on the quality of a friend's outfit, and so on. Physical interactions that were once fire-and-forget events are now being turned into a data memory and can be analyzed indefinitely by intelligent algorithms that can generalize over behavioral patterns, link data to other data, and predict the likelihood of future events to happen. Ethical norms that were once established for the physical world are now being challenged by many kinds of novel situations in the digital world. It is often unclear whether new technologies are only creepy, or that they are possibly illegal, or that they require new laws and regulations. The emerging field of "ethics of algorithms"1 studies the ethical aspects of the algorithmic transformation of our society.  

In this talk I will discuss the ethics of algorithmic developments such as machine learning and artificial intelligence and outline general ethical aspects of algorithms with increasing agency. I will additionally employ the domain of public libraries, and that of archival practices, to illustrate potential effects of digitalization [6] and algorithmization. I will draw upon my research in sensor-based digitalizations of public libraries and the potential of user monitoring and manipulation [3] in the BLIIPS project [1,5,7,8], research into the ethics of acces in archival practices and how human ethical codes of conduct could be utilized to gain insight in how to provide intelligent algorithms with ethical behavior [9], and our recent empirical, qualitative studies with both public library professionals as well as archivists, aimed to map out the perceived threats and opportunities of algorithms in these professions. My aims in this talk are i) introducing the new field of ethics of algorithms and why it is important, ii) illustrating how different algorithms, with different capabilities and levels of agency, create particular ethical issues, iii) zooming in on public libraries and archives as particularly interesting examples of digitalization and algorithmization because they represent vital societal functions, iv) sketching the contours of solutions to ethical challenges induced by algorithms, for example by employing adaptive [2] capabilities of algorithms themselves to learn from human ethical value systems, and v) generally increase algorithmic literacy [4]. 

• [1] http://martijnvanotterlo.nl/bliips.html  
• [2] Wiering, M.A. and van Otterlo, M. (2012) Reinforcement Learning: State-of-the-Art, Springer Berlin Heidelberg (ISBN 978-3-642-27644-6) (Now being translated into Chinese, 2016-2017).  
• [3] van Otterlo, M. (2014) Automated Experimentation in Walden 3.0. : The Next step in Profiling, Predicting, Control and Surveillance in: Surveillance and Society, volume 12, number 2.  
• [4] van Otterlo, M. (2015) The Libraryness of Calculative Devices : Artificially intelligent librarians and their impact on information consumption, in: Algorithmic Life, L. Amoore and V. Piotukh (eds.) Routledge, Taylor & Francis Group, London UK.  
• [5] van Otterlo, M. (2016) Project BLIIPS: Making the Physical Public Library more Intelligent through Artificial Intelligence, Qualitative and Quantitative Methods in Libraries (QQML) 5: 287-300, 2016  
• [6] van Otterlo, M., & Feldberg, F. (2016). Van kaas naar big data: Data Science Alkmaar, het living lab van NoordHolland noord. Bestuurskunde, 29–34  
• [7] van Otterlo, M. (2017) Technical fixes and ethical challenges for activity monitoring in physical public libraries through big data and artificial intelligence --- Invited talk at the 9th International Conference on Qualitative and Quantitative Methods in Libraries (QQML2017) in Limerick, Ireland (May 2017)  
• [8] van Otterlo, M. and Warnaar, M. (2017) Towards Optimizing the Public Library: Indoor Localization in SemiOpen Spaces and Beyond, proceedings of the Benelux Conference on Machine Learning (Benelearn-2017)  
• [9] van Otterlo, M. (2017) From Intended to Intentional Archivists: Ethical Codes for humans and machines in the archives, to appear in Archives in the Infosphere, Smit, F., Glaudemans, A. and Jonker, R. (eds.)  

dr.ir. Martijn van Otterlo obtained his PhD (artificial intelligence, A.I.) from the University of Twente (Netherlands, 2008) with a dissertation on expressive knowledge representation in machine learning from evaluative feedback. He published two books on such adaptive learning algorithms (2009 IOS Press; 2012 Springer, together with Dr. Wiering). Martijn has worked on robotics, vision and language and held positions in Freiburg (Germany), Leuven (Belgium), Nijmegen (The Netherlands). His second research interest, which arose from his expertise in A.I., concerns the ethics and implications of adaptive algorithms on privacy, surveillance and society. He has served as committee member and reviewer for dozens of international journals and conferences on machine learning, data science and artificial intelligence. In his current position at the Vrije Universiteit Amsterdam (The Netherlands) he combines data science and ethics with his third interest: libraries. He currently studies the digitalization of physical, public libraries, and also the ethical consequences of datafication of library processes and services. More information can be found at http://martijnvanotterlo.nl