BBC Research & Development is working with colleagues across the BBC, as well as academic and expert institutions, to develop Machine Learning - and data-enabled technologies more generally - in ways that reflect and uphold core BBC values and support the BBC in delivering its remit.
Project from - present
Why it matters
Machine Learning (ML) has transformative potential for social and public good across a variety of sectors - health, education, media, transport, etc.- but this disruptive potential brings with it a set of societal challenges and raises important questions about the broader social implications and consequences of the technology.
The BBC is currently developing machine learning applications and capabilities and BBC R&D is exploring potential future applications of AI in the media. The risk of unintended societal consequences from ML has been well illustrated by recent high profile cases where ML has been shown to make decisions that are biased or negatively impact on social groups. The opacity of many of these systems further complicates this problem, and there is a lack of clarity in many cases as to where accountability resides in these complex socio-technical systems. The BBC is committed to anticipatory, evaluative and proactive research to advance Machine Learning in the public interest.
Just as our broadcasting and journalism services are built on a number of fundamental principles, based on our public mission (...) the AI services that we build will have these same principles at their heart’’ - Matthew Postgate, 2017.
This programme of work aims to deepen our knowledge of the key challenges facing the media industry, with a specific focus on public service broadcasting to help keep the BBC at the forefront of debates, developments and best practice. Our research agenda aims to develop an approach to ML where BBC values - impartiality, transparency, accountability, independence, universality, and fairness, among others - are embedded and preserved in future development, application and evaluation of machine learning technologies and systems and automated systems more generally.
Current areas of work
- Responsible AI and Public Service Media
- Intelligible AI by design
- Public Service approaches to personalisation and recommendation systems
- Public understandings of AI and attitudes/expectations about use of AI in the media
Following the 2017 BBC conference on Artificial Intelligence and Society, BBC R&D in collaboration with key colleagues across the BBC, conducted scoping work into current debates about ethics and machine learning. We attended a number of key events including The Growing Ubiquity of Algorithms and Society atthe Royal Society the Digital Ethics Summit organised by TechUK and we conducted a comprehensive literature review on the topic. This work culminated in a scoping report on the topic: ‘The case for ethical machine learning at the BBC’. This scoping work informed the BBC’s submission of evidence to the House of Lords Select Committee on Artificial Intelligence and made recommendations for a BBC research agenda to advance work in this area.
These recommendations have now been formalised into the following programme of research:
- Responsible AI and Public Service Media: We are building case studies of ML at the BBC to identify issues and necessary responses to help ensure fairness, transparency and accountability in workflows and systems. We are also supporting academic research into AI, media and bias to inform our work around responsible AI in the public sector.
- Intelligible AI: We are interviewing industry stakeholders about ML and AI systems at the BBC and we are exploring key requirements for explainability.
- Public Service Personalisation: We are investigating approaches to public service recommendations and personalisation that align with BBC values, for example by fostering principles of diversity exposure. This extends to considering new ways to articulate and measure public service value in these systems.
- Audience Research: We are researching audience understandings, attitudes and expectations around automated decisions, ML and the media.
- Convening an internal and external debate: We are working with key people across the wider BBC to convene internal discussion forums and helping our colleagues in the BBC Blue Room and Academy organise the ‘AI, Society and the Media’ conference, including an ‘AI, media diversity’ networking event hosted by the BBC women in STEM network.
How to get involved
This is a pan BBC Research and Development programme of work being done in collaboration between Design and Engineering, and a wider BBC stakeholder group. Our approach is interdisciplinary and collaborative. If you are actively working in this area and you want to share this work with us, or think there might be opportunities to collaborate, we want to hear from you.
This project is part of the Future Experience Technologies section
This project is part of the Internet Research and Future Services section