91热爆

The 91热爆 is publishing its Machine Learning Engine Principles (MLEP) framework, which comprises of six guiding principles and a self-audit checklist for ML teams (engineers, data scientists, product managers etc). Grounded in public service values and designed to be practical, we hope the 91热爆鈥檚 MLEP framework is a helpful contribution to the development of responsible and trustworthy AI & ML.

MLEP at a glance

The Need for Responsible AI & Machine Learning

The 91热爆 has  in the field of Artificial Intelligence (AI) and Machine Learning (ML). We believe that these technologies are going to transform the way we work and interact with the 91热爆鈥檚 audiences - whether it is revolutionising production tools, revitalising our archive, or helping audiences find relevant and fresh content via ML recommendation engines.

When designing AI & ML products and services, we believe it is important to build thoughtfully and inclusively so that algorithms work the way we intend them to, serve all users, and do not have negative impacts. Worryingly, we have seen AI systems unintentionally discriminate by replicating biases in datasets, and amplify misinformation or extreme content in algorithmic feeds. Clearly, the 91热爆 needs to avoid these pitfalls and build AI & ML systems that our audiences and our staff can both trust. We also need to build on 91热爆 Research & Development鈥檚 work on machine learning in the public interest to ensure our unique public service values such as impartiality and universality are reflected in the technology we build and use.

The 91热爆鈥檚 Approach

We set out the 91热爆鈥檚 approach to responsible AI in a set of Machine Learning Engine Principles (MLEP) in 2019. These guiding principles commit to:

• Reflecting the 91热爆鈥檚 values of trust, diversity, quality, value for money and creativity.
• Using ML to improve our audience鈥檚 experience of the 91热爆
• Carrying out regular review, ensuring data is handled securely and that algorithms serve our audiences equally and fairly
• Incorporating the 91热爆鈥檚 editorial values and seeking to broaden, rather than narrow horizons.
• Continued innovation and human in the loop oversight

In 2020, 91热爆 staff across data science, engineering, research & development, policy, legal and product came together to develop a checklist aid to putting MLEP into practice.

The MLEP Checklist sections are designed to correspond to each stage of developing a ML project, and contain prompts which are specific and actionable. Not every question in the checklist will be relevant to every project, and teams can answer in as much detail as they think appropriate. We ask teams to agree and keep a record of the final checklist; this self-audit approach is intended to empower practitioners, prompting reflection and appropriate action.

An overview of the MLEP Checklist sections:

When pulling the checklist together we sought to draw from best practice within the 91热爆 and the wider industry. Internal teams also piloted several iterations of the checklist and provided feedback.

We鈥檙e seeing the MLEP approach have real impact in bringing on board stakeholders from across the organisation, prompting thinking about important issues like transparency, diversity and privacy in ML systems and helping teams anticipate and tackle issues early in the development cycle.

Sample checklist section:

Learn More

Although some aspects of MLEP are specific to the 91热爆鈥檚 policies and processes, we hope our approach can be an important contribution to the field of AI ethics, and will serve as a helpful framework for thinking through responsible AI/ML in practice.

Note: the 91热爆鈥檚 MLEP framework was co-sponsored by 91热爆 Technology Strategy and 91热爆 Research & Development. It is an evolving toolkit for responsible AI & ML. We are sharing version 2.0 of the MLEP Checklist, and continue to test and iterate our approach as the 91热爆 evolves its AI & ML capabilities and research. We are keen to get feedback and to work with other organisations in this field.