Artificial intelligence: How to make fair decisions

All charity professionals have to make decisions about the way their organisations are run. If you’re a charity leader, senior manager, or trustee, the decisions will hold a lot of weight and have far-reaching impact. For example, you may need to make decisions about where funds are spent, who the charity employs, and who benefits from the organisation’s services.

 

In recent years, we’ve seen an increasing number of charities embracing artificial intelligence (AI). AI systems and tools are being used in decision making, such as recruiting staff and making changes in how services are run.

 

This article is part of the Conscious AI campaign which champions fair and responsible use of AI. Below, we explore how charity professionals can avoid bias when using AI in decision making.

 

 

Bias in AI decision making

 

The Charity Commission for England and Wales says that trustees must only make decisions that help achieve their charity’s purpose, so it’s important to consider whether AI supports this requirement. The regulatory body advises that charities should proceed with caution when using AI, as there are risks involved which need to be considered and managed. It explains that “AI is a work in progress so won’t always get things right”.

 

It’s important that all decisions – especially those that have an impact on staff, volunteers, or service users – are fair. The Information Commissioner’s Office (ICO) says that bias is an aspect of decision making which can be detected in AI tools. This is because AI learns from data which may be unbalanced.

 

Bias in decision making can lead to discrimination which can affect people based on their gender, race, age, health, religion, disability, or sexual orientation.

 

Below are some steps for charity professionals to take to avoid bias when using AI in decision making.

 

Understand the impact of using AI

 

Think about how your charity might use AI and whether it’s right for your organisation. This includes looking at the benefits and risks of using AI in decision making. You can do this by carrying out a bias audit and reviewing third-party AI tools and the companies selling them.

 

You should also look at what gaps can be filled or insights generated by an AI tool that you don’t have access to already in your team. Consider your AI-readiness by understanding what skills are needed to use these tools and if existing staff and trustees have those skills.

 

 

Create an AI policy

 

As your AI use develops, you should consider putting together an internal policy to outline where AI is used in decision making. The Charity Commission says to make it clear how and when AI can be used in the organisation’s governance, by employees in their work, and in delivering services. This can help to prevent misuse of AI and the potential for bias in decision making.

 

 

Make sure data inputs are unbiased

 

AI systems are developed and shaped by the data used to train them. These are called data inputs. As business consultancy firm Fisher Phillips says, if data inputs reflect historical bias, inequities and incomplete information, the AI tools and systems may “replicate or amplify those patterns”.

 

One way charities can help to tackle this is by thinking about the data they collect and use. This involves making sure the information they collect isn’t heavily represented or weighted in one area than it should be to reduce bias. When charities rely on biased data to make decisions, the results are often not expected – which could lead to charities making the wrong decisions.

 

 

Checking AI outputs

 

The Charity Commission expects trustees and staff to make sure they have oversight of AI outputs to prevent errors. It says that “the human touch is key to the way many charities operate and interact with their beneficiaries”.

 

 

Regular AI auditing

 

Charities should monitor their use of AI and keep up to date with regulatory and legal requirements. You should consider arranging compliance training for staff and trustees, so they understand their responsibilities when it comes to using AI.

 

 

Removing unconscious bias

 

Grassroots initiative Digital Resistance says that AI bias occurs partly because it learns from data created by humans – who are biased themselves. According to Imperial College London, everyone has unconscious bias, causing us to automatically make quick judgements and assessments on people. This can impact key decisions made in the workplace.

 

Charities should raise awareness of unconscious bias across the organisation, including holding training sessions for staff, trustees, and volunteers. This should cover the importance of not making rushed decisions, recording why important decisions are made, and making decisions as a team.

 

It’s important to have open conversations about bias and how everyone can work together to tackle it. This can help staff and trustees to better manage the output of AI tools in decision making.