Skip to main content
Hafan

Part 2: Explaining AI in practice

Contents

About this guidance

What is the purpose of this guidance?

This guidance helps you with the practicalities of explaining AI-assisted decisions and providing explanations to individuals. It shows you how to:

  • select the appropriate explanation for your sector and use case;
  • choose an appropriately explainable model; and
  • use certain tools to extract explanations from less interpretable models.

How should we use this guidance?

This guidance is primarily for technical teams, however DPOs and compliance teams will also find it useful. It covers the steps you can take to explain AI-assisted decisions to individuals. It starts with how you can choose which explanation type is most relevant for your use case, and what information you should put together for each explanation type. For most of the explanation types, you can derive this information from your organisational governance decisions and documentation.

However, given the central importance of understanding the underlying logic of the AI system for AI-assisted explanations, we provide technical teams with a comprehensive guide to choosing appropriately interpretable models. This depends on the use case. We also indicate how to use supplementary tools to extract elements of the model’s workings in ‘black box’ systems. Finally, we show you how you can deliver your explanation, containing the relevant explanation types you have chosen, in the most useful way for the decision recipient.

What is the status of this guidance?

This guidance is issued in response to the commitment in the Government’s AI Sector Deal, but it is not a statutory code of practice under the Data Protection Act 2018 (DPA 2018) nor is it intended as comprehensive guidance on data protection compliance.

This is practical guidance that sets out good practice for explaining decisions to individuals that have been made using AI systems processing personal data.

Why is this guidance from the ICO and The Alan Turing Institute?

The ICO is responsible for overseeing data protection in the UK, and The Alan Turing Institute (The Turing) is the UK’s national institute for data science and artificial intelligence.

In October 2017, Professor Dame Wendy Hall and Jérôme Pesenti published their independent review on growing the AI industry in the UK. The second of the report’s recommendations to support uptake of AI was for the ICO and The Turing to:

“…develop a framework for explaining processes, services and decisions delivered by AI, to improve transparency and accountability.”

In April 2018, the government published its AI Sector Deal. The deal tasked the ICO and The Turing to:

“…work together to develop guidance to assist in explaining AI decisions.”

The independent report and the Sector Deal are part of ongoing efforts made by national and international regulators and governments to address the wider implications of transparency and fairness in AI decisions impacting individuals, organisations, and wider society.