liveBook · Manning

Data protection impact assessment

As explained in Delivering Machine Learning Projects: From design to deployment

A data protection impact assessment is a structured assessment of how planned personal-data processing may affect individuals and the measures needed to address those impacts.

A data protection impact assessment (DPIA) records how a proposed use of data might affect individuals. Begin once the project hypothesis is clear, while requirements and risks can still be challenged. Treat the assessment as a project artifact, not as a claim that legal compliance alone settles the ethical questions. section 3.5 section 2.6

Write an initial assessment that:

  • Describes the nature, scope, context, and purpose of the data use, including what is new about the proposed processing. section 3.5
  • Checks whether people know about the proposed use and whether consent is in question. section 3.5
  • Reviews possible effects on individuals, such as loss of control, restricted access to services, inability to exercise rights, or financial loss. section 3.5
  • Documents ethical and regulatory requirements, challenges project assumptions, and checks applicable jurisdictional and domain-specific law. section 2.6

Customer purchasing model

A retailer proposes scoring millions of customers for likelihood of stopping purchases, with people—not the model—making decisions from the scores. The data stays under the retailer’s control, and retention and security will not change. Yet using purchasing history to build the model is novel processing. The assessment asks whether customers expect or have consented to that use, and whether offers based on the scores could exclude some customers and cause financial loss. section 3.5

How Delivering Machine Learning Projects: From design to deployment covers it

How deep it goes
Works through assessment questions and an initial customer-data risk review, but does not provide a complete DPIA template.
Examples
Uses customer churn scoring to examine novel processing, consent questions, and possible financial loss from selective offers.
What it assumes
Assumes a defined project hypothesis; requires investigating applicable laws rather than assuming the team already knows them.
What you can do after
Practice/perform an initial assessment of proposed data use and document questions and risks affecting individuals.
How it connects
Connects data processing risks with consent, GDPR, ethical requirements, and algorithmic impact assessment.

Also explained in

Compare how 3 books explain data protection impact assessment →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Delivering Machine Learning Projects: From design to deployment. The section links point to the parts of the book it draws on.