Codes of ethics can be prescriptive (prescribe the do's and don'ts) or aspirational (only specify ideal results). Ferguson et al. (2016) note that they are an important tool for archivists, yet not always sufficient, especially not when there are conflicts between rules and values. Archival codes49 of ethics have a history. The first dates from 1955, from the Society of American Archivists (SAA). It (SAA 1955) is fairly compact and states things like: "The Archivist should endeavour to promote access to records to the fullest extent consistent with the public interest, but he should carefully observe any proper restrictions on the use of records". Similar statements come from the Universal Declaration on Archives (ICA-DL 2011): "Archives are made accessible to everyone, while respecting the pertinent laws and the rights of individuals, creators, owners and users". "The Archivist should respond courteously and with a spirit of helpfulness to reference requests." "The Archivist should not profit from any commercial exploitation of the records in his custody." Later (SAA 1992) it includes: "It is not sufficient for archivists to hold and preserve materials: they also facilitate the use of their collections and make them known." This amounts to the preservation, use and publicity aspects of the archive. It also contains: "Archivists endeavour to inform users of parallel research by others using the same materials, and, if the individuals concerned agree, supply each name to the other party." This refers to a dilemma I have discussed. The final commentary of the code states something about potential conflicts: "When there are apparent conflicts between such goals and either the policies of some institutions or the practices of some archivists, all interested parties should refer to this code of ethics and the judgment of experienced archivists." The most recent version (SAA 2012) features additional core values, which represent what the archivists believe while the code itself represents a framework for the archivists' behavior. This division is intuitive and could be a way to solve some ethical dilemmas, for example by a utilitarian analysis weighing in more factors. For access it expresses the value that it is essential in personal, academic, business and government settings, and use of records should be welcomed. Later in the code of ethics itself this value is translated into "minimize restrictions and maximize ease of access". Ethical codes, especially when they have consequences when misbehaving, cause fewer discipline problems among members (Kizza, 2016, p. 50). However, some codes of conduct can be non-committal. Morris50 calls for an enforceable code of ethics, just like legal and medical professions are governed by codes of ethics which carry the force of law. Violations would then be subject to sanctions including loss of license and civil and criminal liabilities. Formalizing ethical codes though, has one main purpose: to formalize how humans should behave, in this case in the archival profession. I call this the intended archivist; how he is supposed to think, feel and act professionally based on human values and human behavior. By formalizing it in a code it becomes transparent and can be communicated to peers, users, donor organisations and the general public. (4) The Ethics of Algorithms For algorithms, ethical analysis has only started recently resulting in the multidisciplinary field of ethics of algorithms (see for pointers: Lichocki et al., 2011; van Otterlo, 2013,2014a,2014b,2016a; Medina, 2015; Mittelstadt et al., 2016). People often associate with algorithms properties such as infallible, exact, and especially: objective. Because computer-based algorithms are based on logic and statistics people tend to think that because of that algorithms are objective and fair, since they can compute the best answers given the data. While some of this may be true, in general algorithms are far from objective: they are heavily biased (Bozdag, 2013; van Otterlo, 2013). Consider for example51 (part of) a simple algorithm for a bank, specifying that "IF sex female AND age 60 THEN decision no-life- insurance-policy". Now this algorithm is perfectly mathematical, and exact, and it thoroughly computes from personal data whether somebody is eligible for a life insurance policy. However, from a human point of view, it is far from "objective", or "fair" since it discriminates against women above 60 years old. Its decisions are biased and it discriminates, in plain sight. To make things worse, we can also imagine a second algorithm which is specified as "IF f(sex) g(age) 3.78 THEN decision no-life-insurance-policy", and let us assume it makes exactly the same decisions as the first. A problem here is that this algorithm discriminates too, but it is hard to see from its description because we do not know what the functions f() and g() do, and also not why there is a threshold of exactly 3.78. Maybe these aspects have been learned from data which would require us to have a look at the data and learning process to form an opinion about the algorithm's objectiveness. In general, algorithms are biased in many ways (Bozdag, 2013), for example by the data, by learning procedures, by programmers who make choices, by technological constraints and many other reasons. This immediately requires us to form an opinion about algorithms and whether they do the right thing, which again brings us back to ethical reasoning. Characterizing the ethics of algorithms is hard since algorithms and potential consequences are so diverse, and situations may change over time. Mittelstadt et al. (2016) define concerns about how algorithms transform data into decisions, which are then coupled with typical ethical issues. The core operations of an algorithm are: 1) it turns data into evidence which can be a probabilistic prediction, a yes-no decision, or some other conclusion, and 2) it uses the evidence to trigger and motivate an action based on the data. For example, an algorithm for bank loans could take personal data of someone and produce a credit-score of 12, which then could trigger an action to approve a particular mortgage. For the first step three general concerns archives in liquid times 278 martijn van otterlo from intended archivists to intentional algivists. ethical codes for humans and machines in the archives 49 Many other related codes exist, for example by the Dutch royal association for archivists (KVAN)(1) and the professional charter for librarians in public libraries (PL 1993)(2), and codes by the American library organization (ALA)(3), the International Federation of Library Associations and Institutions (IFLA)(4) and the International Council of Museums (ICOM)(5) code of ethics for museums. Although libraries do have different activities, core values are shared with archivists, which can be seen in the similarities with library values concerning access. Occasionally separate codes are made with respect to specific aspects such as privacy, for example as was done recently by IFLA in 2015 (6). See: (1) http://kvan.nl/images/PDF/ Beroepscode_voor_Archivarissen.pdf; (2) http://www.ifla.org/files/assets/faife/codesofethics/netherlands. pdf; (3) http://www.ala.org/advocacy/proethics/codeofethics/codeethics; (4) http://www.ifla.org/news/ ifla-code-of-ethics-for-librarians-and-other-information-workers-full-version; (5) http://icom.museum/ fileadmin/user_upload/pdf/Codes/code_ethics2013_eng.pdf; (6) https://www.ifla.org/node/9803 50 http://slanynews.blogspot.nl/2010/09/enforceable-code-of-ethics-why.html 51 Birkbak and Carlsen (2016) elegantly show in a toy experiment how bias that is explicitly put in (the code of) a ranking algorithm causes different results, exemplifying how implementation choices change algorithm outcomes. As bias, they use intuitive operationalizations of the company mottos of Google, Facebook and Twitter. 279

Periodiekviewer Koninklijke Vereniging van Archivarissen

Jaarboeken Stichting Archiefpublicaties | 2017 | | pagina 141