Data Studies 2021 // S04. Data cleaning

Pablo Velasco // Information Studies // pablov.me

Plan for the day:

• Data taxonomies
  • data is, data can be, data should be: standarisation, uniqueness, outliers
  • Place activity
• Cleaning data
  • bifocalism, data feminism, and how to research cleaning practices
  • Uddannelse activity

DATA TAXONOMIES

• Data can be:
  Kitchin, R. (2014). The Data Revolution: Big Data, Open Data, Data Infrastructures and Their Consequences
  • Quantitative
    • Nominal: categories
    • Ordinal: ranks, scales
  • Qualitative
  • Structure
    • Structured: consistent format
    • Semi-structured: irregular format
    • Unstructured
  • Source
    • Captured: observation, measurement
    • Exhaust: produced by a machine
    • Derived: from additional processing
  • Type
    • Indexical: includes identifiers
    • Attribute: attributes of the identifiers
    • Metadata: data about data

• Data is:
  Floridi, L. (2010). Information: A Very Short Introduction. OUP Oxford.
  

  • Taxonomical: can be ordered or related
  • Typological: primary, derived, metadata

• Data is:
  Rosenberg, D. (2013). Data before the fact.

  • Abstract: an abstraction of "reality"
    • "bracketing unnecessary information from diverse components within a system"
      Cook, W. R. (2009). On understanding data abstraction, revisited
  • Discrete: it consists of finite elements
  • Aggregative: can be combined and accumulated
  • Meaningful: conveys some meaning

"well-formed" data

• Data should be FAIR:
  Wilkinson et al (2016). The FAIR Guiding Principles for scientific data management and stewardship.
  • Findability
  • Accessibility
  • Interoperability
  • Reusability
• Data should be Smart:
  Schöch, C. (2013, November 22). Big? Smart? Clean? Messy? Data in the Humanities.
  • Structured (or semi-structured)
  • Enriched
  • Small

"The chance of mistake is lowered and the chance of success is heightened. The possibility of its continued circulation is better and better for data use." Anne M.

Examples of wild data

• Dataism
  Dijck, J. van. (2014). Datafication, dataism and dataveillance: Big Data between scientific paradigm and ideology.
  • ideology of neutrality
    "Dataism thrives on the assumption that gathering data happens outside any preset framework (…) and data analysis happens without a preset purpose"

Pristine data is often seen as something truthful and valid which make people forget that there lies several invisible processes conducted by other human beings behind this creation/presentation of data." Rune B H

(Click the image for a bigger version. You can also find the gif here)

Short activity:

Discuss with your naboer

1. Identify “things” that you recognise in Place.
2. What kind of things constitute the "preset framework" for place to convey meaning? Is this framework technical, cultural, social?
3. What do you think is the purpose of Place? How does this purpose influences the final product?

CLEANING DATA

• "Process of detecting, diagnosing, and editing faulty data"
  Van den Broecket al. (2005). Data Cleaning: Detecting, Diagnosing, and Editing Data

Most common metaphor for "data scientist" in mediacloud. Image from Catherine D'Ignazio. Data from www.mediacloud.org. (Data feminism ch.5)

• 80% of data analysis spent cleaning and preparing data
  Dasu T, Johnson T (2003). Exploratory Data Mining and Data Cleaning.

Data, capta, sublata

"It is very much akin to cleaning silverware; one needs to make the surface presentable without damaging the delicate nature of such an object." Mikael SA

• Data (given) != capta (taken)
• Not passive acceptance, but active construction
  Drucker, J. (2011). Humanities Approaches to Graphical Display
• Knowledge as produced, more than discovered
  Gitelman, L. (2013). Raw Data Is an Oxymoro

"If we say it’s just given, that worries me a bit because it means you just seize and receive it. But in fact you actually elicit it, and eliciting is what I think this term sublata captures".
Latour, B. (2017). Gaia or knowledge without spheres

• Aesthetic practices:

  • data practices that elicit, extract, and select, rather than merely produce data.

• Standarisation with consequences:
  • "do not simply make data transparent and open but enable comparison of the relative performance of populations as objects of knowledge and governing."

Ratner, H., & Ruppert, E. (2019). Producing and projecting data: Aesthetic practices of government data portals.

https://pudding.cool/2018/08/pockets/

• Data Feminism:
  D’Ignazio, C., & Klein, L. F. (2020). Data Feminism

  • "a way of thinking about data, both their uses and their limits, that is informed by direct experience, by a commitment to action, and by intersectional feminist thought."
  • "how race, class, sexuality, ability, age, religion, geography, and more are factors that together influence each person’s experience and opportunities in the world. Finally, data feminism is about power - about who has it an who doesn't"

• "openness is not singular in its effects but has different consequences for the populations that are enacted"
  Ratner, H., & Ruppert, E. (2019). Producing and projecting data: Aesthetic practices of government data portals.

Facebook gender fields circa 2014. Photo by Lauren F. Klein (Data Feminism ch.4)

Bivens, R. (2017). The gender binary will not be deprogrammed: Ten years of coding gender on Facebook.

• Bifocalism of knowledge:
  • sites of production != sites of projections

Latour, B. (2017). Gaia or knowledge without spheres

How to observe cleaning practices?

• "Showing all the variations in processing and making public all the ways processors “work differently” would, it was feared, under-mine the impression of standardized internal processing."
  Plantin, J.-C. (2019). Data Cleaners for Pristine Datasets: Visibility and Invisibility of Data Processors in Social Science.
• observe the role of automation (question tools and techniques)
• notice instances of bifocal practices (e.g. "forced blindness")

How to observe the complexity of existing datasets?

https://uddannelsesstatistik.dk/pages/gymnasialeudd.aspx

"This is a source of data friction, as the schools cannot simply change their reporting of numbers without it having consequences for their own planning purposes (…) Rather than changing the 'school reality', the aesthetic practice of data cleaning resolves this friction so that data can be projected in the forms required by the warehouse"
Ratner, H., & Ruppert, E. (2019). Producing and projecting data: Aesthetic practices of government data portals.

Short activity

Go to https://uddannelsesstatistik.dk/Pages/Reports/1693.aspx and look for information on your own gymnasium

1. Can you find any information on the collection or processing of data?
2. Is there any information on the "quality indicators"?
3. Where does the "socioøkonomiske referencer" metric come from?
4. Do you think is possible to objectively compare all gymnasiums in Denmark?