Site navigation

Leader Insights | Transforming the BBC with Responsible AI & ML

Ross Kelly


Responsible AI
DIGIT spoke with James Fletcher, Responsible AI Lead at the BBC to discuss how the broadcaster’s community-driven ethical framework shapes its approach to AI innovation.

In recent years, the development and use of artificial intelligence and automated systems has been the source of much contention. Concerns over privacy, data bias and the potential discriminatory impact of artificial intelligence are a hotly-debated subject, prompting heated discussions on ethical and responsible development.

Despite lingering doubt and concerns, artificial intelligence continues to play an increasing role in business transformation, enabling organisations to streamline processes, improve customer experience and deliver personalised services.

James Fletcher, who leads the responsible application of AI at the BBC, says the situation is no different at the public broadcaster. In 2017, responding to the House of Lords Select Committee on Artificial Intelligence, the BBC pledged it would play a key role in cultivating “the right environment for AI” in the UK.

As part of this commitment, the BBC announced it would help to inform the debate surrounding artificial intelligence through its editorial function while facilitating closer collaboration between “public service institutions, academia and the commercial sector” to examine the key issues in this space.

Crucially, the BBC said it would place a keen emphasis on the responsible development and deployment of artificial intelligence and use machine learning to “enrich users’ lives”.

Since this pledge, the broadcaster has accelerated its use of artificial intelligence to enhance services, features and curate content for millions of viewers, listeners and readers each year.

Alongside cutting-edge research happening in the BBC’s R&D department, a key player is the Datalab team, which uses machine learning to provide users with personalised recommendations and drive content discovery across several BBC platforms.

“The Datalab team is involved with BBC Sounds, iPlayer and the Sport and News product teams, helping improve the relevance of content we suggest to users,” Fletcher explains.

“The BBC creates a large amount of content every day and has a very diverse audience, so trying to find the right piece of content for individual users is a challenge.”

Keeping pace in a changing world

Although a strong focus is placed on audience-facing machine learning efforts, Fletcher says the broadcaster also views artificial intelligence as means by which it can improve production processes and support staff in their day-to-day tasks.

“In a public facing sense, this is about improving our outcomes with audiences and their engagement with our products or services. The BBC’s agenda is all about delivering value for all our audiences,” he says.

Internally, the organisation is examining how it can improve the efficiency of its broader production activities, viewing machine learning as a vital tool to “augment human creativity, skills and capabilities”.

“This is about liberating their creativity by automating and replacing some of the more mundane tasks staff deal with each day,” he explains.

BBC R&D has a range of projects underway, exploring new ways to support staff across a number of functions. Machine learning is being trialled by editorial teams to find impactful images for news articles, and to source images for programmes in iPlayer.

Meanwhile, computer vision is being leveraged to detect and classify animals caught on camera by teams working on the Springwatch series. This has helped drastically cut the time spent sifting through raw footage for a nugget of eye-catching content.

“We’re freeing up staff from some of the tedious tasks,” Fletcher explains.

“These technologies are acting as a multiplier in many ways. With Springwatch, that led producers to discovering content that would’ve otherwise gone left unseen, or storylines that wouldn’t have ended up in the programme.

“Using these gave the time back to producers to focus on the storytelling and focus on the creative side of things,” he adds.

These examples provide a mere snapshot of the variety of work underway at the BBC, Fletcher notes. Crucially though, they offer an insight on how the BBC views the development of data-driven services as it traverses its way through an increasingly congested media landscape.

“Our ability to keep up is very important,” he says. “We create so much content, so making sure that our audiences and users are seeing is relevant is super important to us.”

“It’s not just about keeping up, though. It’s about doing it in a distinctive BBC way. Our approach to recommendations, as an example, will not be the same as other players. We want to do this in our own way which encompasses our public service values.”

Responsibility and Obligations

As Britain’s public broadcaster, Fletcher says the need to keep pace and continually innovate raises unique challenges with regard to ethical and responsible use.

In his role at the BBC, he is tasked with ensuring AI/ML is harnessed in a manner that delivers tangible benefits to the organisation while considering the person on the other side of the screen.

“In my role, it’s about making sure that what the BBC does with AI and ML is aligned to the BBC’s values and principles,” he explains. “As a public service organisation, the BBC’s charter requires it to do what it does in a responsible way.”

Having been around for nearly 100 years, this public service culture is woven into the very fabric of the organisation. And notably, it is this culture that plays a key role in steering its approach to the adoption of new technologies.

Projects utilising technologies such as ML or AI work within the Machine Learning Engine Principles (MLEP) framework. These help ensure BBC teams are building ML in a responsible way and in line with BBC values.

“It’s meant to be a tool to inspire thinking,” Fletcher explains, with its initial development a community-driven effort by staff from a number of functions across the organisation.

“This came from a very strong, passionate community of people at the BBC who’ve been pushing this for a while. They’ve been pushing both the BBC’s experimentation with the adoption of these technologies, but also promoting the idea that we should be doing it in a responsible way,” he says.

Read more Leader Insights

Fletcher is keen to emphasise that the MLEP framework is intended as much more than just a box ticking exercise. Running parallel to these principles, the BBC has developed a checklist and guidance for teams to consider throughout the entire process of scoping, building and deploying machine learning systems.

“The idea is that this should be in people’s minds from the very beginning and inform them every step of the way. So teams are considering this from the start – from discovery through to deployment and then also afterward, helping to monitor and evaluate when a project is live,” he explains.

Since its inception there have been a number of revisions to the MLEP framework, and the community continues to play a key role in its future direction. It’s not intended as a one-size-fits-all guide, and Fletcher says the next step is to continue to make it more flexible to accommodate for the wide range of projects that could be in-play at any given moment.

Looking ahead, this also provides a strong foundation for the governance and audit of BBC ML systems, Fletcher says, and will be updated and refined to ensure it remains fit for purpose.

“The idea is that this will evolve as technology is moving fast in this area, plus our appreciation of the underlying issues is also moving rapidly. I see MLEP as a service or a product. And as such, it needs to be continually changed and iterated.”

While the negative implications of artificial intelligence often dominate public discourse, there are more positive examples of AI application coming to the fore, and the BBC’s ongoing work in this space shows the potential for organisations to innovate responsibly and at scale.

Furthermore, in taking a community-led approach to help steer this transformation, the team has demonstrated a key route to help shape organisational alignment on a process fraught with challenges and ethical considerations.

DIGIT Expo 2021 | Join the Conversation

James Fletcher will explore how the BBC effectively leverages machine learning at DIGIT Expo 2021 on 23rd November, sharing additional insights on responsible AI development and the Machine Learning Engine Principles framework.

DIGIT Expo is Scotland’s largest gathering of senior IT and Digital personnel. With 1000+ delegates, 40+ speakers and 50 exhibitors, the event is an unmissable opportunity for knowledge exchange, networking and business opportunity.

For information on how to register, visit:

Ross Kelly

Staff Writer

Latest News

%d bloggers like this: