1 of 47

Features of the Data Landscape

(Big) Data & Data Assemblages

Ceilyn Boyd

ceilyn_boyd@harvard.edu

ceilyn.boyd@simmons.edu

NEASIST

11 January 2019

2 of 47

Navigating the Data Landscape - What features will you encounter?

2

3 of 47

Connecting data practice and research

3

Practice

Harvard Library

Research

Simmons University

(Big) Data &

Data Assemblages

?

Critical Data Studies & Philosophy of data

Research Data Management

4 of 47

Concepts

4

5 of 47

5 Key Concepts

(Big) Data

Measurements, pre-factual, pre-analytical pieces of information exhibiting 3Vs (volume, variety & velocity)

Assemblages

Non-hierarchical arrangements of components characteristized by extrinsic properties and relations

Data Assemblages

Sociotechnical infrastructures concerned with data

Research Data Management

Active management of data throughout the research and data lifecycle

Critical Data Studies

Interdisciplinary research area concerned with the critical, systematic investigation of data & data assemblages

5

6 of 47

Term: Critical Data Studies (CDS)

Interdisciplinary research area concerned with the critical, systematic investigation of data & data assemblages

6

7 of 47

Critical Data Studies (CDS)

7

Data

big data

Critical Theory

Discipline

small(er) data

Critical Data Studies

Bias
Ethics
Justice
Power
Privacy
Technology

data assemblages

Social Theories

assemblage theory

Geography

(or LIS, RDM, etc.)

8 of 47

Foundation of Critical Data Studies

Origins

Geography, sociology
Circa 2012
Against Dataism
boyd, danah, & Crawford, K. (2012). Critical questions for big data: Provocations for a cultural, technological, and scholarly phenomenon.

7 Guiding Principles

Data are situated in time & space
Technology is not neutral
Data is shaped by context
Data is not raw
Data complements other ways of knowing
Data should be used in progressive ways
Scholars (geography, LIS) should engage in praxis, not solely theory

(Dalton & Thatcher, 2014)

(Oliphant, 2017)

8

9 of 47

Term: (Big) Data

9

10 of 47

Data are not pure or natural objects with an essence of their own. They exist in a context, taking on meaning from that context and the perspective of the beholder.

(Borgman, 2015, p. 18)

10

“

11 of 47

Big Data possess a suite of key traits: volume, velocity and variety (the 3Vs), but also exhaustivity, resolution, indexicality, relationality, extensionality and scalability...Our analysis reveals that the key definitional boundary markers are the traits of velocity and exhaustivity.

(Kitchin & McArdle, 2016, p. 1)

11

“

12 of 47

(Big) Data Characteristics

Data

(Usually) Operationally defined

Context-dependent: data when, not what

Many formats (e.g. structured to unstructured)
(Often) Machine-readable
Not raw or neutral

Big Data

Data and datasets that exhibit some or all of the following properties:

Volume
Velocity
Variety
Exhaustivity
Resolution & Indexicality
Relationality
Extensibility & Scalability

(Kitchin & McArdle, 2016, p. 2)

12

Here are some characteristics of generic data and of Big Data.

Data

None of the data characteristics are surprising. But I'd like to call out the last two: machine readable and not-neutral.

In our library and archival collections, we have a lot of data that is not yet machine-readable and/or machine-actionable -- what I call computationally accessible.

It's trapped in an analog format and cannot easily be manipulated with computers. It's good to remember these cases because just because data is (computationally) accessible doesn't mean that it's all the data on the topic that exists!

Example

For instance, I'm thinking of a particular collection at the Harvard University Archives, a collection of Colonial-era diaries written by John Winthrop that contain direct numerical and qualitative astronomical and climate observations. The documents have been digitized--so they are machine readable, after a fashion, but they are not computationally accessible in the form of a spreadsheet, for instance. We have other collections with trapped data, like this one: records of financial transactions and inventories, ships log books with weather and tide data.

It's good to be reminded that, although data can appear to be highly objective, it is a source of potential bias--it is not neutral. We know that its uses may not be neutral (as has been shown with social media data, most recently). But the measurements themselves may be biases.

Big Data characteristics

Volume: quantity (e.g. number of Gigs)
Velocity: frequency of updates
Variety: diverse formats, types
Exhaustivity: all of the data, not just a few samples
Resolution: fine grained
Indexicality: able to identify a specific individual
Relationality: able to be connected/composed with other datasets
Extensibility: data fields can be expanded
Scalability: data may rapidly increase in size/velocity/some other dimension

The size of the data and its shape has implications for the kinds of infrastructure and policy needed to secure, store, and manipulate it.

My research area--Critical Data Studies--tackles these issues of data epistemology and non-neutrality head-on.

Reference

Kitchin, R., & McArdle, G. (2016). What makes big data, big data? Exploring the ontological characteristics of 26 datasets. Big Data & Society, 1–6.

13 of 47

Data: Small & Big

13

Characteristic	Small Data	Big Data
Volume	Limited to large	Very large
Exhaustivity	Samples	Entire populations
Resolution & Indexicality	Coarse and weak, to tight and strong	Tight and strong
Relationality	Weak to strong	Strong
Velocity	Slow, freeze-framed, or bundled	Fast, continuous
Variety	Limited to wide	Wide
Extensibility & Scalability	Low to medium	High

(Kitchin, 2014, p. 28)

(Kitchin & McArdle, 2016, p. 2)

This slide shows some key characteristics of Big Data and small data

Often, data are only big or small in comparison to one another and depending upon the organizational, management, and computational tools that you have at your disposal

The existence of Big Data might fool people into thinking that analyses based upon it is more stable/rigorous than it really is. Big Data may be biased, or skewed and certain populations may not be represented in the data

Small Data as Context for Big Data

One key thing to note is that, often small data provides context for Big Data. And that just having Big Data is not enough--it's not the End of Theory (https://www.wired.com/2008/06/pb-theory/ and https://onlinelibrary.wiley.com/doi/full/10.15252/embr.201541001)

The implications and importance of combining Big Data with small data is one research area within critical data studies.

References

Ford, H. (2014). Big Data and Small: Collaborations between ethnographers and data scientists. Big Data & Society, 1(2).

Kitchin, R. (2014). The data revolution: big data, open data, data infrastructures & their consequences. Los Angeles, California: SAGE Publications.

Kitchin, R., & McArdle, G. (2016). What makes big data, big data? Exploring the ontological characteristics of 26 datasets. Big Data & Society, 1–6.

Welles, B. F. (2014). On minorities and outliers: The case for making Big Data small. Big Data & Society, 1(1).

14 of 47

(Bigger) Data in a (Harvard) Library Context

14

Characteristic	Examples
Volume	Digitized & digital collections Born digital collections Electronic records Scientific research data
Exhaustivity	Scientific research data
Resolution & Indexicality	Scientific research data
Relationality	Collections as Data Research data
Velocity	Research data
Variety	Born digital collections Digitized collections Metadata Research data
Extensibility & Scalability	Born digital collections Collections as Data Research data

15 of 47

Term: Assemblage

Non-hierarchical arrangement of components characteristized by extrinsic properties and relations

15

16 of 47

An assemblage is an arrangement of heterogeneous, autonomous components whose extrinsic properties and interactions give rise to a unique, enduring individual.

(Boyd, 2018, p. 5)

16

“

17 of 47

[T]he assemblage permits the researcher to speak of emergence, heterogeneity, the decentred and the ephemeral in nonetheless ordered social life.

(Marcus & Saka, 2006, p. 101)

17

“

18 of 47

Assemblage

Definition

A real, non-hierarchical, nested arrangement or system of heterogeneous, autonomous components defined by exterior relations

Origins

Social theory, political philosophy, realist tradition
Assemblage Theory - Deleuze & Guattari (1980)
Neo-Assemblage Theory - DeLanda (2006, 2016)

Purpose & Use

Describe micro- and macro-aspects of social reality
Widespread use in social sciences & CDS, literature, very limited use in LIS

Related Concepts & Theories

Rhizome, Network, Actor-network theory (ANT)

18

An assemblage is a particular type of arrangement (or system) of heterogeneous components

An assemblage has a specific type of topology, just like a network does, or a tree or a directed acyclic graph

Very limited use in LIS - some in archives (rhizome) some in library management.

Rhizome: "(Gilles) Deleuze and (Felix) Guattari use the terms "rhizome" and "rhizomatic" to describe theory and research that allows for multiple, non-hierarchical entry and exit points in data representation and interpretation."

Social scientists use the assemblage concept to describe and critically examine complex, dynamic, and heterogeneous phenomena across a variety of disciplines.

Primary goal is to describe the micro- and macro-aspects of social reality.

Assemblage theory is widely used in social science research because of its expressive power; the affordances it provides for the critical analysis of power relations; and its ability to be combined with other social theories, such as critical theory and Actor-Network Theory (ANT).

I will show you a simpler definition and description of assemblage theory later in the presentation

19 of 47

Assemblage Characteristics

Assemblages are...

Real
Heterogeneous
Non-hierarchical
Tri-scale (3): spatial, temporal, nesting
Dynamic
Exhibit emergent behavior
Composable
Machine-like
Members of populations
Responsive to internal & external events

Assemblages have...

Components
Component interactions & relations
Capacities - respond to events
Processes - cohesion, dissolution
Properties
Possibility space
Roles (3) - material, expressive, hybrid
Scope

19

As you can see from the list of characteristics, assemblages are very conceptually dense. They pack a lot of concepts and relationships under one umbrella term: Assemblage.

They have smooth, machine-like behavior and upon first glance, they may appear like seamless wholes. But on deeper inspection, they reveal levels of internal nesting and interaction with other assemblages.

Animation 1

The composable and machine-like quality is why I have appropriate the gear icon to describe assemblages.

So one question arises, why use them to describe phenomena? Why not use something simpler?

The answer is, that in everyday contexts, when you just want to describe something, using a more simple concept, like a network, may be fine.

It is primarily in a research context, when we need to capture very specific nuances in order to do fine-grained comparisons that assemblages become very useful.

Assemblage theory gives us a lexicon to talk more precisely about the behavior of complex, real world organizations and their interactions with one another.

20 of 47

Examples of Real-World Assemblages

20

Here are some real world examples of assemblages

Animation 1

Physical sciences: Galaxy (contains planets, stars, asteroids, comets);

Animation 2

Molecule (contains atoms); Atom (contains protons, neutrons, electrons)

Biological sciences: Cell (contains mitochondria, cilia, cytoplasm); Animal (contains organs, bones, blood);

Animation 3

A city, with buildings and roads, people, bridges, cars and trucks

Other examples include:

Nations, with states, inhabitants, laws, and land

References

DeLanda, M. (2006a). A new philosophy of society: assemblage theory and social complexity. London: Continuum.

DeLanda, M. (2006b). Deleuzian Social Ontology and Assemblage Theory. In M. Fuglsang & B. Meier Sorensen (Eds.), Deleuze and the Social. Edinburgh University Press.

DeLanda, M. (2016). Assemblage theory. Edinburgh: Edinburgh University Press.

21 of 47

21

22 of 47

Sesay, A., Oh, O.-O., & Ramirez, R. (2016). Understanding Sociomateriality through the Lens of Assemblage Theory: Examples from Police Body-Worn Cameras.

22

23 of 47

Diagram of an assemblage that emphasizes its machine-like, interlocking qualities

(Boyd, 2018)

23

24 of 47

Term: Data assemblage

Sociotechnical infrastructure concerned with data

24

25 of 47

Data Assemblage

Definition

A sociotechnical infrastructure concerned with data

Origins

Critical Data Studies (CDS), circa 2014

Purpose & Use

Examine social dimensions (e.g. power relations, cultures) associated with sociotechnical infrastructures
Critical data studies, Critical algorithm studies

Related Concepts & Theories

Assemblage, Network, Platform (platform studies), Actor-network theory (ANT), Infrastructure theory (e.g. Star & Ruhleder, 1996)

25

26 of 47

Data Assemblage Characteristics

Data assemblages (are)…

Embody and encode standards
Encodes knowledge
Invisibly support use
Reflect systems of thought (e.g. dataism, capitalism)
Sites of data practice & culture
Sources and enablers of (big) data

Embedded in other infrastructures (e.g. technical, policy, financial)
Are learned as a part of membership

Data assemblages have…

Components that become visible upon systems breakdown
Intersections with social, economic, and governmental entities

26

27 of 47

27

28 of 47

28

29 of 47

Here’s How #Ferguson Exploded on Twitter Last Night (Oh, 2014) https://www.motherjones.com/politics/2014/11/ferguson-twitter-map

29

30 of 47

Amazon Doesn’t Consider the Race of Its Customers. Should It?

(Ingold & Soper, 2016)

https://www.bloomberg.com/graphics/2016-amazon-same-day/

30

31 of 47

Research

Build infrastructure to systematically describe, analyze & compare data assemblages

31

32 of 47

32

Me

CDS

Analyze & Compare

Data assemblages

Critical data studies is a small, but growing area of research.

The fallout from the 2016 election, and the role that social media has played in reproducing and reinforcing propaganda and influencing public opinion and political behavior, its vulnerability to bad actors, has increased scrutiny of data assemblages and research in the area.

To date, there are a lot of theory pieces and 'calls to arms' pieces that have been written. But scholars such as Rob Kitchin, have called for more empirical studies. Particularly more studies that compare data assemblages.

There are a lot of challenges to conducting studies, for example:

Gaining access to sites of data practice
Expense and complexity
Lack of tools for standardized analyses

This is a gap in the literature …

Animation 1

...and an opportunity for me to make a contribution to the discipline.

Animation 2

33 of 47

Overview of Research Direction

Research Questions

How can we systematically describe, analyze, and compare assemblage phenomena?
How can assemblage theory be made more accessible and useful to LIS researchers?
How can we support empirical studies involving data assemblages?
How can we design better data assemblages?

Approach

Develop domain ontology & glossary for assemblage theory
Develop propositions & conceptual framework for assemblage theory
Extend infrastructure to include data assemblages & conduct empirical studies
Use infrastructure to inform system design and data modeling (MVC -> OVC)

33

34 of 47

Knowledge Organization Systems - Ontology is most complex

Pieterse, V., & Kourie, D. G. (2014)

34

Q: How can we systematically describe, analyze, and compare assemblage phenomena?

A: Build an ontology & glossary

35 of 47

A. Workflow to Develop Domain Ontology & Glossary

35

Statistics	#
Total concepts	71
Top-level concepts	8
Descendents for each top-level concept	36
assemblage	6
capacity	3
event	0
mechanism	1
population	1
process	5
property	20
virtual diagram	0

36 of 47

Visualization of Assemblage Theory Domain Ontology

36

37 of 47

Domain Ontology & Glossary

Domain Ontology

OWL File (XML) & Class documentation (PDF)
https://bit.ly/2Fksu3U

Glossary

Textual Glossary (PDF)

https://bit.ly/2H1TPtV

37

38 of 47

B. Workflow to Develop Assemblage Theory Conceptual Framework

38

Q: How can assemblage theory be made more accessible and useful to LIS researchers? How can we support empirical studies involving data assemblages?

A: Build a conceptual framework

39 of 47

B. Assemblage Theory Conceptual Framework Summary

8 Theory Propositions

Real
Heterogenous
Non-hierarchical
Tri-scale
Dynamic
Has possibility space
Belongs to population
Composable

12 Concepts

Assemblage
Capacity
Component
Event
Mechanism
Population
Process
Property
Role
Scale
Scope
Virtual diagram (possibility space)

3 Relationships

Is-a
Has
Member-of

3 Facets

Causality
Probability
Structure

39

40 of 47

Assemblage Theory Conceptual Framework Diagram

40

Assemblage

Probability

Causality

Structure

Virtual Diagram

Event

Mechanism

Process

Virtual diagram*

Capacity

Population

Virtual diagram*

Assemblage

Component

Property

Role

Scale

Scope

Virtual diagram*

41 of 47

41

42 of 47

Overview of Research Direction - Redux

Research Questions

How can we systematically describe, analyze, and compare assemblage phenomena?
How can assemblage theory be made more accessible and useful to LIS researchers?
How can we support empirical studies involving data assemblages?
How can we design better data assemblages?

Approach

Develop domain ontology & glossary for assemblage theory
Develop propositions & conceptual framework for assemblage theory
Extend infrastructure to include data assemblages & conduct empirical studies
Use infrastructure to inform system design and data modeling (MVC → OVC)

42

43 of 47

Summary

43

44 of 47

Why Assemblage Theory & (Data) Assemblages?

Assemblage concept is conceptually dense, complex, but powerful
Assemblage-like phenomena are ubiquitous, relevant to many disciplines & practices
Assemblage Theory can be paired with other theories (e.g. LIS--information behavior) to enrich research and practice
Critical Data Studies

Opportunities for critical engagement with data within LIS and RDM

Ex. Library as site of data practice; Data literacy; Data justice & advocacy
Ex. Navigation, use, and design of RDM assemblages

44

45 of 47

Questions?

Thank you.

ceilyn_boyd@harvard.edu

ceilyn.boyd@simmons.edu

45

46 of 47

References

Borgman, C. (2015). Big data, little data, no data: scholarship in the networked world. Cambridge, Massachusetts: MIT Press.

boyd, danah, & Crawford, K. (2012). Critical questions for big data: Provocations for a cultural, technological, and scholarly phenomenon. Information, Communication & Society, 15(2), 662–679.

Boyd. C. M. (2018). A Prototype Domain Ontology for Neo-Assemblage Theory. Manuscript in preparation.

Dalton, C., & Thatcher, J. (2014). What does a critical data studies look like, and why do we care?

DeLanda, M. (2006a). A new philosophy of society: assemblage theory and social complexity. London: Continuum.

DeLanda, M. (2006b). Deleuzian Social Ontology and Assemblage Theory. In M. Fuglsang & B. Meier Sorensen (Eds.), Deleuze and the Social. Edinburgh University Press.

DeLanda, M. (2016). Assemblage theory. Edinburgh: Edinburgh University Press.

Ford, H. (2014). Big Data and Small: Collaborations between ethnographers and data scientists. Big Data & Society, 1(2).

Kitchin, R. (2014). The data revolution: big data, open data, data infrastructures & their consequences. Los Angeles, California: SAGE Publications.

Kitchin, R., & McArdle, G. (2016). What makes big data, big data? Exploring the ontological characteristics of 26 datasets. Big Data & Society, 1–6.

Marcus, G. E., & Saka, E. (2006). Assemblage. Theory, Culture & Society, 23(2–3), 101–106.

Oliphant, T. (2017). A case for critical data studies in library and information studies. Journal of Critical Library and Information Science Studies, 1(1).

Welles, B. F. (2014). On minorities and outliers: The case for making Big Data small. Big Data & Society, 1(1).

46

47 of 47

Credits

Presentation template by SlidesCarnival
Photographs by Unsplash, Pexels
Icons by:

47