1. What are standards?

 

What sort of standards is the AI Standards Hub focused on?

 

Our focus in this introduction – and the focus of the AI Standards Hub – is on standards produced by recognised Standards Development Organisations (SDOs), sometimes referred to as ‘technical standards’. SDOs are standard-setting organisations that follow certain principles [link to process section] and are often formally recognised in international treaties and regulations.

 

Standards of this kind can be defined as documents developed through consensus-building processes and approved by a recognized body. They provide common and repeatable rules, guidelines and characteristics for activities or their results, and are aimed at providing order in each context.

 

Throughout this introduction, we will be using the term ‘standards’ rather than the more specific term ‘technical standards’ which is sometimes used when discussing standards developed by SDOs.

The term ‘technical standards’ is commonly used in policymaking and academia to help make clear their distinction from regulatory, professional, industry and other types of standards. However, the term is not widely used beyond these circles. When we engaged with stakeholders, we found that in broader usage, the term often causes confusion since standards of the relevant kind can take a variety of forms [link to kinds of standards section] many of which are not strictly ‘technical’ as the name suggests.

 

Standards and standardisation

 

Standards are essentially ‘agreed ways of doing things’. The point of a standard is to provide a reliable basis for people to share the same expectations about a product, process, system, or service. Standards have two important characteristics: they provide a guide for behaviour, setting the expectations that organisations should adhere to; they also provide a guide for judging that behaviour, allowing people to understand whether their expectations have been met, enabling them to place their trust wisely. (For more details on how standards are developed, see section 5 [link])

 

Standards address a wide spectrum of issues that are relevant to our everyday lives. These range from specifying the technical requirements that enable us to connect to the internet via WIFI networks, to setting out guidance on corporate social responsibility and environmental sustainability.

 

Some standards are written such that an external party can check whether they have been adhered to, through a conformity assessment procedure. Independent assurance providers can test and certify the conformity of products, processes and services against standards, should an organisation seek it. 

 

The result of a conformity assessment can be reported publicly, to demonstrate trustworthiness and build public trust. Conformity can also be communicated through labels such as BSI’s Kitemark, the UK’s UKCA mark, or the European CE mark which demonstrate that a product conforms to the relevant regulations.

 

Aside from high-risk contexts such as the manufacture of medical devices, external assessment and certification is not a requirement for declaring conformity with a standard. Organisations can also declare their own compliance with certain standards, via a conformity attestation.

 

Participation in the development and use of standards is voluntary (though they can have important links to regulation see section 3 [link]). This voluntariness is a defining characteristic of the kinds of standards the AI Standards Hub focuses on, influencing how these standards are developed as well as the shape of the resulting standards.

 

Because the use of standards is voluntary, they need to be established in a way that ensures that they can gain wide acceptance, mainly by organisations that will use them but also from the individuals and organisations who will be affected by them. If standards are not seen to be convenient or beneficial to users, or society more broadly, they simply would not be used.

 

For this reason, the standards development process – known as standardisation – is a process of consensus building. The standards that result from these processes express what parties involved have managed to agree upon. An important aspect of the standards development process is that it provides a forum for building common ground, though understanding the needs of different contexts and societies, and addressing and resolving tensions.

 

There are limits to what can be standardised, particularly on a global level. For risk management, for example, we can standardise the way that we describe different risks. What we cannot do is standardise what level or types of risk are acceptable or unacceptable at a global level – these are inherently political, contextual decisions. The task for standardisation is to address issues in a way that can provide uniform, practical guidance, while leaving the space for different jurisdictions to make these decisions.

 

Different types of standards

 

Standards are used across a broad range of sectors and application areas. They can be used to codify technical specifications regarding the measurement, design or performance of products and systems, as well as the impacts or efficiency of broader processes or services.

 

For example, standards can be as straightforward as providing simple metric definitions, such as the Ö2:1 ratio that underpins the standardised format of A4 and related paper sizes that have been adopted in many countries across the world. But standards can also provide guidance on complex, context specific processes such as managing organisations’ environmental and societal impacts and responsibilities.

 

As a result, there are many different types of standards, written for a variety of purposes, including:

 

Setting common terms and definitions

 

Foundational and terminology standards make sure that everyone can speak the same language, enabling effective communication between diverse stakeholders. They help to build a common language by setting out agreed-upon terms and definitions. These standards also provide the foundations on which other standards can be built.

 

Defining processes and approaches

 

Process and management standards define organisational processes and approaches. For example, providing repeatable guidance for risk management procedures or outlining factory best practices to ensure that the food that ends up in our supermarkets is safe to eat.

 

Standardising test methods and performance requirements

 

Product testing and performance standards set out repeatable methodologies and requirements to test the quality, safety and performance of products and processes. For example, testing medical devices and determining the accuracy thresholds at which devices can be deemed safe for use in healthcare.

 

Defining metrics and reference points

 

Measurement standards set the ultimate points of reference upon which measurement equipment, testing and performance standards can be built.  

 

Setting out requirements for compatibility and interoperability

 

Interface and networking standards ensure that products and systems can work together, increasing the scope and usability of products and systems, and reducing waste and duplication.