There is always a social machine which selects or assigns the technical elements
used' (Deleuze Parnet, 2007, p. 70). This is an important notion which might be
helpful to disentangle the sometimes-confusing relationship between data and
records.
In a world of ubiquitous computing the ability to define data-points and to monitor
and record data has become infinite. CISCO expects that in 2020 more than 50
billion devices are connected to the Internet and these devices 'require minimal
human intervention to generate, exchange and consume data' (Rose cs, 2015,
p. 17). In the past, it was a time-consuming human activity to select the elements
worthwhile to be recorded. That process, the 'conscious or unconscious choice
(determined by social and cultural factors) to consider something worth archiving'
was coined by Eric Ketelaar as archivalisation (Ketelaar, 1999). Archivalisation
precedes archiving and to understand this process, we need to understand what
Hofstede called the 'software of the mind' which is programmed by social and
cultural factors and comes very close to Deleuze's 'social machine'. Nowadays all
particles that are 'observed' by a machine are recorded, although, and that is not
unimportant, the data points to be monitored and recorded still need to be defined
and programmed. The recorded raw data in itself is meaningless; these are signals
without real significance. Data are only meaningful in relationship with other data,
processed in a specific situation. Only from that perspective the concept of the
record or archive is a meaningful construct. It means that archives should be seen as
Foucauldian apparatuses (of governance), dispositives, machineries of seeing, but,
and that has to be emphasised, machineries of seeing from a particular point of view
(Giannachi, 2016, p. xv-xvii; McQuillan, 2016, p. 8). Thinking about the archive via
the apparatus means focusing on the networked arrangement of media,
mechanisms of communication and data processing (Packer, 2010). The archive is
meaningless without understanding the interdependency of these socio-technical
components and humans. Only then we will be able to understand, as Geoffrey
Bowker writes, that every act of permitting 'data into the archive is simultaneously
an act of occluding other ways of being, other realities. The archive cannot in
principle contain the world in small; its very finitude means that most slices of
reality are not represented' (Bowker, 2014, p. 1797, italics CJ). One could argue that,
compared to the past practices of recording, the number of potential witnesses
within a situated practice have incredibly increased by the explosion of sensors and
data-points. Nevertheless, what is represented by records is in the end based on
situated needs, that define the technical arrangements.
Back to appraisal
I opened this article with Calvino's quandary how to distinguish between the
meaningful and meaningless. I conclude with the proposition that in our time of
ongoing datafication of society, archivists need to redefine the record and as a
consequence of it the recordkeeping mechanism of distinguishing between the
essential and residue. I showed that protecting informational accountability
requires rethinking the components of the record or archive. The apparatus-view, in
which the archive functions as a machine of governance, is helpful to understand
the intricate, assemblage-based relationality between the components of the archive.
The recorded data is only one element of that machine. What data is relevant, and
which other components are required to create a meaningful record, is defined by
the situated context of operation. The Volkswagen emission scandal of 2015 may
serve as an example. Advanced software in the diesel engines of Volkswagen could
detect when the car was tested and subsequently adapt its emissions during the
artificial test circumstances to acceptable levels. Back on the road, the vehicles
switched to normal mode with much higher emission rates. It is an example that
shows that the recorded data of the tests can only be understood in combination
with the software, and that the software can only be understood if the logic of the
algorithms is known. If the record only provides the recorded data, it is not sufficient
for informational accountability. Above all, algorithms are models written with a
specific purpose and they are not neutral nor objective. Understanding the results of
algorithmic processing requires at least knowledge of the underlying assumptions of
the model and the data which are used by the algorithms. The familiar principle of
'the context is all' is also applicable in this layer of construction of the record.
This has major implications for the issue of appraisal, which gets a much wider scope
than just answering the question of keeping or discarding recorded data. From a
recordkeeping perspective, the issue is not about data; it is about what people,
institutions and communities want to be able to reconstruct for purposes of
business, evidence, accountability and memory. That perspective is decisive for
answering the question which components of 'the archive as an apparatus' should
be preserved in coherence. Providing robust accountability is not an easy and
especially not a pure technical task to accomplish. Joshua Kroll, who developed a
general framework for accountable algorithms in automated decision-making
processes, stresses that accountability requires the possibility to verify that 'the
social, legal and political structures that govern an automated process function as
they are intended to function' (Kroll, 2015, p. 210). Robust accountability requires
involvement in the system design and computer systems should be designed in a way
that they are reviewable (Kroll, 2015, p. 188-202). Cathy O'Neil started a business
to audit algorithms. In an interview with the Los Angeles Times, she explains 'I don't
want to just audit a specific algorithm by itself, I want to audit the algorithm in the
context of where it's being used. And compare it to that same context without the
algorithm' (Los Angeles Times, 2016). Nicholas Diakopoulos argues that a new
accountability perspective is necessary in freedom of information requests.
Although there are some examples of successful use of Freedom of Information Act
requests to compel disclosure of source codes (Diakopoulos, 2016, p. 59), this is
definitely not sufficient to guarantee accountability. He suggests reconsidering FOIA
along the lines of Freedom of Information Processing Act which is not so much
based on disclosing codes, but allow to submit benchmark datasets the government
agency is required to process through its algorithms (Diakopoulos, 2016, p. 59).
These are just some examples of efforts to accomplish informational accountability.
Does this imply a profound reorientation of the archival community? Yes and no.
No, since it is all about understanding the context. But the efforts to be made to
understand the context of creation and use require a reconsideration of the
components of the record. The archival community needs to rethink and
reconceptualise the essence of a record in a world in which data is ubiquitous, fluid
and too abundant to manage and control. If the archival community wants to
continue to play a meaningful role in defending informational accountability and
archives in liquid times
206
charles jeurgens threats of the data-flood. an accountability perspective in
the era of ubiquitous computing.
207