Parliament Resolution on Robotics, debated in 2017. Article 12 of this resolution
says that it should always be possible to supply the rationale behind any decision
taken with the aid of Artificial Intelligence (AI) that can have a substantive impact
on one or more persons' lives. It must always be possible to reduce the AI system's
computations to a form comprehensible by humans. Interestingly that same article
articulates the necessity that 'advanced robots should be equipped with a 'black box'
which records data on every transaction carried out by the machine, including (my
italics CJ) the logic that contributed to its decisions'.8 This is in line with the
regulation on the protection of natural persons with regard to the processing of personal
data and on the free movement of such data which was adopted in April 2016 by the
European Union. This regulation sets rules and requirements for data-driven
automated processing. Every person has 'the right not to be subject to a decision
which is based solely on automated processing (...)'.9 Examples that are explicitly
mentioned are automatic refusal of an online credit application and e-recruiting
practices without human intervention. Predictive profiling, one of the most
fundamental intrusions in a private life which is increasingly used by authorities in
fighting crime and terror remains allowed 'where expressly authorised by Union or
Member State law'.
Big data analysis is useful for revealing general patterns, but, and that is not always
kept in mind, there is always a probability of a mismatch between general patterns
and a specific situation (WRR, 2016, p. 27). Scholars, journalists, advisory and
legislative bodies warn against excessive techno-dependency and techno-optimism.
The Dutch investigative journalist Dimitri Tokmetzis criticises the naïve way rules
are formulated and used in algorithms without paying enough attention to the
validity of the underlying assumptions. To give an example: Dutch government
assumes that poverty is a risk factor for the education of children. It is possible to
design an algorithm to find evidence of poverty in the electronic child records and to
make a list of families that need to be watched closely to be able to intervene if
necessary. But do we know whether the assumption behind the rule is valid? Is the
assumption that defines the rule based on thorough scientific research? (Tokmetzis,
2012, p. 59-60). Scott Mason, researcher at Keele University, also warns against the
often-careless way how bureaucrats and policy-makers interpret and contextualise
the results of Big Data itself without consulting domain experts to assess the validity
of correlations. He is very critical about the claim that Big Data analysis creates the
possibility for 'neutral' evidence based policy-making. According to Mason, 'the vast
quantities of correlations generated by Big Data analytics act simply to broaden
the range of 'evidence from which politicians can chose to support their arguments'
(Mason, 2016). In 2016, the Dutch Scientific Council for Government Policy
notified a highly undesirable tendency of policymakers to accept the revealed
patterns without questioning the validity of the results for specific situations
(WRR, 2016).
Archival implications
The aforementioned examples show that computational algorithms increasingly
become an integrated part of government processes and decision making. Legal
scholars have argued for more than 20 years in favor of more transparency in
automated processing (Kroll, 2015, p. 6). Since records which are created in the
course of business, 'provide evidence of actions, decisions, and intentions, both
legal and illegal, proper and improper, and wise and misguided' (Cox Wallace,
2002, p. 4), availability of records is vital for accountability and transparency. Also
in the recordkeeping realm, algorithmic tooling is used to manage the growing
number of documents and to find relevant information for a specific purpose. The
most advanced developments of algorithmic computation in the recordkeeping
sphere can be found in eDiscovery and information retrieval applications.10
Since archivists claim to play a pivotal role in defending institutional and societal
transparency and accountability (Jimerson, 2009, p. 246-252), there is an urgent
need for archivists to ruminate what it means to take this role in the era of
ubiquitous computing. If records, archives and archivists want to continue to be
key players in ensuring and defending accountability, this evokes the question what
meaningful recordkeeping is in this new context of data-ubiquity, and at the same
time what meaningful records are.
As Upward and others have put forward, this is exactly one of the main challenges
the archival and recordkeeping community is confronted with: to clarify how the
conceptual relationship between data, records and archives is designated in the era
of ubiquitous computing. The traditional record was based on fixity and stability in a
material sense. What has fundamentally changed is the possibility to produce
different aggregates out of the same recorded data, which means, as Bruno Latour
(2009) writes, 'that the whole has lost its privileged status' which makes us aware of
the fact that the whole is always simpler than the parts (p. 198). The written record
used to have the shape of an entity (the whole) in which the parts (words, sentences,
paper, lay out, signature etc.) were a fixed materialised aggregate. Since the
computational turn, it is possible to use the same recorded parts (data) in different
configurations simultaneously. The stable whole has been replaced by a 'continually
evolving liquid assemblage of action' (Introna, 2016, p. 19). It is as if we construct
different types of houses with the same bricks at the same time.
This (informational) fluidity is an important feature of what was designated by
Deleuze as an assemblage. An assemblage is in the words of Deleuze 'a multiplicity
which is made up of many heterogeneous terms and which establishes liaisons,
relations between them [T]he assemblage's only unity is that of co-functioning'
(Deleuze Parnet, 2007, p. 69). In an assemblage, an element can be dissociated
from a specific assemblage and continue to function in another assemblage.
Assemblages exist merely because of the relationships between the elements.
Deleuze emphasises that an assemblage is never technological: '[t]ools always
presuppose a machine, and the machine is always social before being technical.
archives in liquid times
204
charles jeurgens threats of the data-flood. an accountability perspective in
the era of ubiquitous computing.
8 European Parliament, Motion for a European Parliament resolution. Report with recommendations to
the Commission on Civil Law Rules on Robotics, A8-0005/2017, art. 12, available at
<http://www.europarl.europa.eu/sides/getDoc.do?pubRef=-//EP//TEXT+REPORT+A8-2017-
0005+0+D0C+XML+V0//EN> accessed at 30 March 2017.
9 Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of
natural persons with regard to the processing of personal data and on the free movement of such data, and
repealing Directive 95/46/EC (General Data Protection Regulation), art 71, available at <http://eur-lex.
europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32016R0679&from=EN> accessed 30 March 2017.
10 The Information Governance Initiative Community provides an interesting overview of activities in these
fields: http://iginitiative.com/community/
205