| Competition | Previous Entries & Ideas | Digital Collections | TOCs | FAQs | Judging | Resources and Tools | Submit Entry | Events |

Kevin Walker, Kate McLean and Sam McElhinney

Generating stories from archives using agent-based simulation

Submitted Entry for 2013 Competition


We aim to generate narratives from digital archives which are spatialised in the form of hypothetical labyrinths, and autonomous agents who 'smell' their way through them to retrieve simple stories. The accumulated trails of numbers of agents would begin to illuminate previously undiscovered links between objects in the form of who, what, where, when, why and how. This could apply to a wide range of digital collections using keyword search of full text or metadata, and we aim to test using the BL's electronic journals database.

Assessment Criteria

The research question / problem you are trying to answer

Please focus on the clarity and quality of the research question / problem posed:

Can autonomous agents create useful and interesting stories from spatial representations of digital archives, using a novel model of smell perception?

Autonomous agents are defined below; 'useful and interesting' stories is a necessarily personal, subjective metric which we will look to BL curators and readers to evaluate.

Please explain the ways your idea will showcase British Library digital collections

Please ensure you include details of British Library digital collections you are showcasing (you may use several collections if you wish), a sample can be found at http://labs.bl.uk/Digital+Collections

Our idea could apply to a number of BL collections, especially digitised manuscripts and publications with full-text search. A particular research interest of ours is research data - therefore the Electronic Journals database would provide a suitable use case.

Please detail the approach(es) / method(s) you are going to use to implement your idea, detailing clearly the research methods / techniques / processes involved

Indicate and describe any research methods / processes / techniques and approaches you are going to use, e.g. text mining, visualisations, statistical analysis etc.

We aim to design and evaluate knowledge construction using digital archives by the following means:

1. Spatialising the archive. Mapping search results, a subset or entire archive exploits proximity and relationships of data, facilitates pattern recognition, and invites exploration. The space we have in mind would take the form of a dynamically reconfigurable maze; imagine an architectural floor plan or city map into which we can insert barriers or release 'smells' to influence agent behaviour.

2. Autonomous agents are released into the space. These agents have information-seeking behaviour based on their physical context (location, characteristics of the space, nearby objects); social context (other agents they communicate with either co-located or remotely); and personal context (simple rules which govern their search behaviour, as well as their history, preferences and 'personality'). Agents navigate using a smell model in which they become aware of nearby objects based not on vision but on location, intensity, range and direction, which we have used previously to describe smells in urban environments. Agents can additionally leave their own olfactory trails, which could then influence other agents' behaviour.

3. The labyrinth itself acts as curator, knowing the best trails through objects. It contains some rudimentary intelligence to introduce attractors or barriers to influence agent behaviour. Additionally, each object has a basic knowledge of its own personal context, when an agent visits it and how many agents have visited it overall.

4. Agents return with stories. These arise from sequences of objects visited in the spatialised archive, and follow simple journalistic story construction rules of who, what, where, when, why and how. 'Who' can include other agents encountered, or arise from the content of the objects they encountered (names attached to database entries for example). 'What' denotes objects encountered and any relevant agent behaviour. 'Where' and 'when' can be simple location timestamps. 'Why' relates to agents' motivation - the rules that govern their search behaviour. And 'how' denotes the route they took to reach each object; this is important with regard to the spatialisation of data, as routes may be direct to varying degrees, so this is about the experience of the search - the blend of factual object data and fictionalised space and search behaviour - though the data could not be mapped onto actual spaces.

It should be noted that all of the above can apply to either computational or human agents. In the real world, an exhibition is an effective spatialisation of archive data in which human agents visit objects of interest in particular physical, social and personal contexts, following particular paths and (ideally) producing stories as a result. Though our project is focused on digital simulation in the first instance, we would love to evaluate agent-based story construction at human scale in the British Library (with or without actual smells), and will be happy to work with BL curators on this.

Please provide evidence of how you / your team have the skills, knowledge and expertise to successfully carry out the project by working with the Labs team

E.g. work you may have done, publications, a list with dates and links (if you have them)

Dr Kevin Walker leads the new Information Experience Design programme at the Royal College of Art, which focuses on experimental design and research between the digital and physical worlds. He is also a Visiting Fellow at London Knowledge Lab. He also designs installations and software for museums, as consultancy (clients have included Centre Pompidou, the V&A, various galleries); and formerly as Senior Software Designer for exhibitions at the American Museum of Natural History. He holds a PhD in Museums and Technology, an MPS in Interactive Telecommunications, and BA in Anthropology and Mass Communications. Further details of his practice is at http://www.walkerred.com, research and publications at http://lkl.ac.uk/kevin.

Kate McLean is Senior Lecturer in Graphic Design at Canterbury Christ Church University, and a sensory researcher, designer and photographer. Her award-winning smell maps of cities around the world have been exhibited at Milan Design Week and in various galleries and group exhibitions, and featured on BBC, Fast Company and Smithsonian. She holds an MFA in Graphic Design, BA in Related Arts, and is currently a PhD candidate at the Royal College of Art. Her work can be seen at http://www.sensorymaps.com.

Sam McElhinney is a principal of MUD Architecture and tutor at UCL and UCA, as well as a Project Manager/Designer at Jason Bruges Studio. Previously he was a key member of Surface Architects, an award winning design practice. He holds Masters degrees from the Bartlett, UCL, and Cambridge. He won the the Ambrose Poynter Prize for his thesis 'Labyrinths, Mazes and the Spaces Inbetween'. His work on agent-based simulations can be seen at http://www.mudarchitecture.com.

Please provide evidence of how you think your idea is achievable on a technical, curatorial and legal basis

Indicate the technical, curatorial and legal aspects of the idea (you may want to check with Labs team before submitting your idea first).


Kevin and Sam are currently working on agent-based simulations of a museum exhibition for UCL's Design with Heritage project, for comparison with actual visiting data. We believe our idea is easily achievable using simple keyword searches of BL datasets, and even more so with access to metadata and database tables. Simple (dumb) agents can be easily implemented, and we plan to experiment with increasing their 'intelligence' by assigning them rules governing preferences and enabling them to learn from experience. They can be given 'personality' by assigning them personal 'likes and dislikes' and by adjusting the tone of voice of the stories they produce.


Kevin's PhD was on visitor-generated trails in museum exhibitions, which this project seeks to evaluate in agent-based simulations. Depending on the collection chosen for the initial study, we would seek curatorial assistance in identifying relevant objects, metadata and content.


We do not forsee legal issues regarding use of the technologies suggested, and defer to the BL on aspects related to the collections. We are not aiming to track human visitors, so privacy issues are not relevant.

Please provide a brief plan of how you will implement your project idea by working with the Labs team

You will be given the opportunity to work on your winning project idea between July 6th - October 31st 2013

We envision working with BL curators and Labs team primarily at the start (July) and end (Oct) of the project.

Provisional plan as follows:
July: Choose collection, test spatialisation & mappings, plan & test agent rules
Aug-Sept: Agent simulation iterations, analysis of agent behaviour & stories generated
Oct: Refine mapping, agent behaviour & story construction with curators & Labs team; initial dissemination of results.