Introducing The Big Reveal (WT)

Tim Hopkins

Project aims

How can uses of AI and adaptive technology be used to create narrative fiction in audio form, responsive to location, in AR experiences for audiences using headphones and mobile devices?

Imagine an audience member at point A, wearing headphones, listening to pre-scripted narration that reveals creative spirits at large – talking to listeners, drawing them into a story.  Other scripted text awaits at locations B, C, D etc.

The user moves unpredictably between them – as they do, AI generates spoken text that bridges the story from one place to another, regardless of the route, with compelling narrative traction.   The thresholds between AI and not-AI may be undetectable, or may announce themselves, like doors to a different realm…

The Big Reveal (WT) is a project researching how to make this possible, bringing together colleagues from different disciplines: Tim Hopkins, Sam Ladkin, Andrew Robertson, Kat Sinclair and David Weir.  We’ve had very welcome support also from Jo Walton and other colleagues, and have arranged future contribution from Victor Shepardson (voice synthesis.)

New developments in generative pre-trained transformer (GPT) technology extend uses of deep learning to produce text. This is a branch of machine learning/AI that has many potential uses and societal impacts.  The project explores this using AI and AR as a creative space for collaboration between engineers and artists.  

We are interested in affording practitioners across these disciplines a space to learn by making something together, perceiving AI as a new space for creative writing in a multidimensional expressive space, and making something that might offer the public an engaging way to experience and reflect on their space and the growing presence and impact of AI.

The project envisages three elements, each with its own development stage.

1) a text-generation system that is adaptive to context in a way that sustains / plays with suspension of disbelief

2) voice synthesis that can translate text into convincing speech / narrative voices

3) a platform combining software which can fuse detection of user activity and location with adaptive delivery on a mobile device

Progress so far

This has focused on 1 (as this will define scope for an application for a large project grant supporting the research phases 2 and 3).

Our method has been to develop a lab simulation combining some key technology, functionality and artistry as a proof-of-concept.

We come from different disciplines.  One of the inherent stimulations for the project is navigating differences between how we conceive and share what we are trying to do.  For example, David and Andrew (Informatics) have provided essential insights and guidance on current work with language models, to help induct Tim, Sam and Kat (MAH) into a complex and huge field.   T, S and K have experience of writing / creating for related spaces (e.g. games, adaptive systems, branching narratives), as well as more traditional contexts, but the concepts and engineering potentially underpinning our project ask for new understandings.  Similarly, discussions of language features at work in creative writing (e.g. complex implications of syntax) may test the functionality and limits of existing automated language models.

A central task has been to look for an approach that can respond significantly to what might be minimal input (prompts) from the user.  In contrast to some game formats, where players’ active choices overtly steer subsequent events, we are interested in an experience where users might not perceive any instructional dialogue with a system at all, but feel as if they are being told or are immersed in a narrative that recognises what experience they are having, and is able to determine what they should hear next.  This needs to happen according to a given narrative world and its bespoke generative language model – adapting to information detected by a mobile device as to location, orientation, direction, speed.

A series of discussions (April-November 2022) each led to tests based on existing text2text approaches, whereby text is put into a language model that infers suitable responses based on a range of data.  Although ultimately in our user’s experience there may be no apparent text prompt from the user themselves, there is nonetheless a need for an underlying element of this kind in order to stimulate a generated response. ‘Text’ in this case may be adaptive written text users actually hear, or an equivalent input determined by their behaviour, generating ‘text’ or prompts that may be hidden from users’ perception.  Our tests involved texts / prompts written by Andrew, Kat, Tim and Sam, fed through a number of text generation processes (on , a prominent platform for AI projects.)

Instead of shorter prompts leading to consequent longer responses, many of these processes were originally designed to achieve different kinds of results – such as inferring summaries of information from input texts.   This tended to result in short outputs that were not quite right for our purposes in a variety of ways. Extending prompts did not flex responses more effectively. Varying the character of prompts, for example imitating strongly-flavoured genres, had some perceivable impacts, but not decisively.  We needed to develop functionality towards richer responses.  This suggested adjusting our approach, involving two current directions.

Next steps

Firstly,  we continue to explore core needs – text2text generation, and training a GPT2-like model. However, we’re focussing on getting a ‘good start’ (DW) to an automated response of the kind we might want – rather than concerns about the length of response (which can be addressed later.)  We are also identifying specific corpora to fine-tune a model. Andrew has been experimenting for example using ‘film reviews’ as inputs (recently using Kat is supplying examples of poetry (including her own) and shortly larger corpora needed to train a classifier – something that can distinguish between kinds of input in the way we need.  Andrew is now working on building a test model based on GPT2 to be fine-tuned with this.

Secondly,  the creation of some kind of ranking machine.  For example

a) we give the same input to create a range of candidate texts (e.g.100), a machine ranks them according to merit, and randomly chooses from the top of pile

b) we have two blocks of text – one visible one not.  We insert candidate third blocks between the visible and the hidden, and rank the insertions according to how well they work between the two. (This discussion also included ‘similarity metrics’ and BERT representation – ‘Bidirectional Encoder Representations from Transformers’).

c) we compare prompts with two corpora of texts – one has features we are looking for (e.g. of genre or form), the other is overtly more neutral (e.g. informational website like BBC news) – the machine ranks highest those closest to the first.

In the new year (2023) we will pick up on these, aiming to make our proof-of-concept model in the Spring.   Towards our grant application, we will also start scoping Phase 2 – on voice synthesis – with input from Victor Shepardson (Iceland, Dartmouth,  delayed this year due to COVID19 impacts.) We will look at challenges and consequences for translating text responses into persuasive speech simulation, and practical issues around processing – since the outcome envisages accompanying users across ‘borders’ in real time, between recorded adaptive narration and AI assisted/generated narration.

What is Concept Analytics and who are we?

Concept Analytics Lab (CAL) gathers linguists, AI engineers, and historians and is aligned with Sussex Humanities Lab within the Critical Digital Humanities and Archives research cluster. The principle mission behind Concept Analytics is to understand human thinking by analysing conceptual layering in texts. We overcome the divides between humanities, AI, and data science by harnessing the power of computational linguistics without losing sight of close linguistic analysis. 

Although CAL was formally set up in 2021, its existence is the culmination of research energies over the previous few years and our desire for a stable space to explore concept-related ideas with like-minded scholars.  Establishing the Lab has provided us with a platform from which we showcase our research expertise to researchers and other external partners. CAL has grown and changed through 2022, during which time we have counted on a team of six researchers at a range of stages in their careers, from undergraduate to postdoctoral level. The team is led by Dr Justyna Robinson. 

CAL has so far partnered with research groups within Sussex, e.g. SSRP, as well as ones further afield, e.g. Westminster Centre for Research on Ageing. We have worked closely with Archives such as Mass Observation Archive and Proceedings of the Old Bailey, as well as non-academic organisations. 

What were the highlights of the past year? 

Our activities in the past year centred around exploring the content of the Mass Observation Project (MOP) and their Archive of May 12 Diaries with the aim of identifying conceptual changes that happened during Covid-19. We have completed two main research projects. CAL was awarded funding through the UK-RI/HEIF/SSRP call Covid-19 to Net Zero, in collaboration with industry partner Africa New Energies, to identify the impact of Covid-19 on people’s perceptions and habits in the context of household recycling and energy usage. CAL was also commissioned by the PETRA project (Preventing Disease using Trade Agreements, funded by UKPRP/MRC) to discover key themes and perceptions the public holds towards post-Brexit UK trade agreements. Keep reading for summaries of the findings of these research projects, as well as our other achievements this year. 

Household recycling with Africa New Energy (ANE) 

Through this project we identified that respondents to the MOP Household Recycling 2021 directive were deeply committed to recycling, but that these feelings were coupled with doubt and cynicism in relation to the effectiveness of the current system. MOP writers pointed to a perceived lack of transparency and standardisation in recycling processes and systems. Lack of transparency and standardisation have also been identified as obstacles to recycling adherence and efficacy in more policy-based analytical surveys (Burgess et al., 2021; Zaharudin et al., 2022). Changes in recycling habits among the UK population were identified as resulting from external factors, such as Covid-19 and reduced services, as well as lack of knowledge about how and what can be recycled. This research has significantly impacted the way our grant partner ANE approach their operations in terms of gaining energy from organic waste content. The research results also led ANE to start work on gamifying the waste classification process. It aims to encourage recycling compliance by replacing the current sanction-based system with a more rewards-based system. This research shows that the CAL already has a track record of establishing commercial routes of impact for our research and we see extending the scope of this impact to be a critical next step in CAL’s research programme. Further details on the collaboration with ANE can be found in this blog post.  

We are seeking further HEIF funding to expand on the work already done with the Household Recycling directive to maximise policy impact by processing the handwritten answers and also processing the 2022 12 May Diaries for insight into the impact of the current energy crisis on respondents’ behaviour and attitudes to energy. As part of this project we would hold an exhibition in which we would invite various stakeholders including policy makers to showcase our work. 

MOP UK Trade Deals 

We were commissioned by the PETRA project’s lead Prof. Paul Kingston from the University of Chester to perform a conceptual linguistic analysis of the MOP UK Trade Deals directive. We used our approach to identify hidden patterns and trends in the answers to the directive questions. The conceptual analysis allows us to combine quantitative with qualitative methods and identify otherwise unperceived patterns. The main themes that arose were related to the perceived quality of trade deals and concerns about animal and ethical standards. We also performed an analysis linked to people’s knowledge, belief and desires. The results of the analysis will inform policy makers in their decisions regarding trade deals. Additionally this piece of work has attracted some interest from public health bodies with whom we are preparing a potential grant for future research. 

Papers and presentations 

In 2022 Justyna Robinson and Julie Weeds both presented the work they did within the context of the Old Bailey archives and have had their paper on that work published in the Transactions of Philological Society. In this paper they describe a novel approach to analysing texts, in which computational tools turn traditional texts into a corpus of syntactically-related concepts. Justyna Robinson and Rhys Sandow also have authored a paper forthcoming in 2023, ‘Diaries of regulation: Mass Observing the first Covid-19 lockdown’. This research will be presented at Mass Observation’s 85th Anniversary Festival, Mass Observation Archive, The Keep, 23rd April 2023. 


As part of the SSRP/HEIF funding we received earlier this year we have also developed a website, which can be found at, where we also post blogs with news pieces and short research insights.