We are committed to working with research leaders to gain the most impactful scientific insights from our data. We wish to harness their cutting edge scientific approaches to provide novel findings that may help us improve our services and that may be of benefit to the wider sector. In particular, we are developing partnerships with researchers across several fields of expertise, including mental health, psychology, psychiatry, neuroscience, computer science including data science and artificial intelligence, linguistics, and public involvement and engagement.

As for many scientific problems, we believe that great breakthroughs are made when researchers from different fields are brought together. This is particularly important in mental health research involving large datasets that require computational expertise to be interrogated effectively - consequently we are keen to build cross-disciplinary collaborations as our work progresses.

We currently have academic partnerships with the Institute of Global Health Innovation at Imperial College London, as well as Imperial’s EPSRC Centre for Mathematics of Precision Healthcare, and Data Science Institute.

We are continuing to develop partnerships with researchers and institutions across several fields of expertise, including mental health, psychology, psychiatry, neuroscience, computer science, linguistics, and public involvement and engagement.

At the launch of Shout 85258 in May 2019, we announced the beginning of our first academic research partnership, with the Institute of Global Health Innovation (IGHI), at Imperial College London, led by Lord Ara Darzi. We have since broadened the partnership to include scientists in Imperial's EPSRC Centre for Mathematics of Precision Healthcare, and Data Science Institute. A number of projects are now ongoing, across several teams with complementary expertise, with a focus on using Natural Language Processing and machine learning, including deep learning models, to understand Shout 85258 conversation content at scale. These are state-of-the-art approaches for computational 'understanding' of language, which is a key challenge with respect to analysing our dataset, which includes many hundreds of thousands of text-based conversations.

Early signs are that these approaches can be used to identify, with striking accuracy in some cases, topics of conversation (e.g. suicide, self-harm, anxiety), demographics of texters (e.g. age, gender), and stages of the conversation itself (e.g. building rapport, problem-solving). These findings are already helping us paint a more detailed picture of the texters that use Shout 85258 and what they contact us about. Building on this work we are seeking a deeper understanding of how these deep neural networks learn to predict these features of conversations (explainability) and what that can tell us about the texters and their mental health.

As this research progresses, we are committed to publishing the findings in peer-reviewed journals for anyone to read and also to ensure that the results are communicated widely in other accessible formats for a range of audiences. Moreover, as our dataset grows, so too do our research ambitions. We are currently developing a number of academic partnerships with other institutions to expand the breadth and depth of expertise brought to bear on the challenge of using our data to better understand mental health.