Museums and heritage institutions (MHIs) are increasingly using AI tools such as Machine Learning, Natural Language Processing, and Machine Vision to enhance visitor interaction with their collections. A well-established problem with AI is bias, including how AI algorithms reproduce skewed underlying data. The inaccuracy of facial recognition AI when applied to people of colour is well known example. For MHIs, a challenge for responsible AI use lies in how underlying biases in museum collections, such as those rooted in colonial knowledge and power, are reproduced through AI data processing and outputs. When seeking to publicly communicate stories of, from, and through their collections, and when using colonial collections as the 'data' for AI mechanisms that facilitate this, what might 'responsible AI' look like for MHIs? What might be the implications, challenges, and opportunities for responsible AI? The project has been developed in consultation with The Royal Armouries, the project's external partner. The Royal Armouries are the UK's national collection of arms and armour and one of the oldest museums in the world. The museum's focus on the history of war forms a backdrop for interpretations encompassing science and technology, politics, law, art and poetry, and the wider human experience. The collaboration connects to ongoing work The Armouries are undertaking that relates to the colonial character of their collections, and to transform their digital offering for visitors. The project will work with this partner and other MHI stakeholders to ensure knowledge outcomes have wide applicability. The project's objectives are to: 1) Scope the terrain of the context and application: map how AI has and could be used to enhance visitor experience; consider the colonial form and origin of underlying data within many museum collections; generate understandings of the actors, stakeholders, interests, and power relations that are relevant to the use of AI to communicate colonial museum collections. 2) Explore how responsible AI can be defined in the setting and application: how using underlying data from a museum's colonial collection in AI presents challenges and opportunities; explore how AI use might reproduce but also complicate colonial forms of knowledge within visitor experiences. 3) Engage with a wider community of MHIs and industry stakeholders to assess whether existing approaches to responsible AI can be used or developed for the particular challenges and opportunities of the context and application. 4) Work with these communities to develop and propose novel responsible AI tools and practices where necessary to assist MHI use of AI. The project will yield a Toolkit, a report, and an academic journal article (detailed below). The research will benefit all MHIs who are considering utilising AI to communicate colonial collections and also yield knowledge outcomes relevant to the broader sector regarding how to navigate bias within collections and archives when utilising AI. The project will present concrete knowledge outcomes encouraging responsible development of AI tools in a museum setting, helping MHIs address the character and origin of collections and more effectively tell their many stories.

A key factor in reducing potential gun crime is to detect someone carrying a gun before they can commit a criminal act. This detection can be achieved by the existing, and widespread, CCTV camera network in the UK. However, the performance of operators in interpreting CCTV imagery is variable as they are trying to detect essentially a very rare threat event. Additionally, current automated systems for detecting possible anomalous behaviour have been found to have varying success. We propose the development of a new machine learning system for the detection of individuals carrying guns which will combine both human and machine-based factors. Using selected CCTV footage which depicts people carrying concealed guns, and other control individuals, the proposal will establish what overt and covert cues (essentially conscious and subconscious cues) experienced CCTV operators actually attend to when identifying potential gun-carrying individuals from such CCTV imagery. In parallel, a machine learning approach will establish the machine recognised cues for such individuals. The separate human and machine cues will then be combined to form a new machine learning approach which will be fully tested. The system will be capable of learning and reacting to local gun crime factors which will aid its usefulness and deployment capability.

Sound, Craft, Vision, Place will draw together a team of researchers from the arts and humanities to assist the promotion, development and provision of innovative outreach activities to encourage community groups to explore and articulate their heritage for themselves, and to stimulate projects for HLF funding. The project themes emphasise the importance of oral history, music, art and design, and digital media at the University of Huddersfield, and our ability to draw on expertise beyond the humanities to provide tools that will attract community interest and prompt project ideas. Community groups themselves will be encouraged to specify the kinds of research areas and methods where collaboration or help would be welcomed. We are especially keen to involve those who have encountered barriers to enjoying the heritage, and an important part of our project will be the role to be played by local and virtual communities in the interpretation of their own backgrounds and surroundings, including web- and digital-based activity to encourage virtual volunteering as a communal activity, undertaken both by people who are connected by place and/or by shared experiences and interests. Among our partners will be the National Coal Mining Museum for England, English Heritage, the Royal Armouries, as well as local community groups, with all of whom we shall be working to develop ideas for projects involving free access to archive materials.

A key factor in reducing potential gun crime is to detect someone carrying a gun before they can commit a criminal act. This detection can be achieved by the existing, and widespread, CCTV camera network in the UK. However, the performance of operators in interpreting CCTV imagery is variable as they are trying to detect essentially a very rare threat event. Additionally, current automated systems for detecting possible anomalous behaviour have been found to have varying success. We propose the development of a new machine learning system for the detection of individuals carrying guns which will combine both human and machine-based factors. Using selected CCTV footage which depicts people carrying concealed guns, and other control individuals, the proposal will establish what overt and covert cues (essentially conscious and subconscious cues) experienced CCTV operators actually attend to when identifying potential gun-carrying individuals from such CCTV imagery. In parallel, a machine learning approach will establish the machine recognised cues for such individuals. The separate human and machine cues will then be combined to form a new machine learning approach which will be fully tested. The system will be capable of learning and reacting to local gun crime factors which will aid its usefulness and deployment capability.

A key factor in reducing potential gun crime is to detect someone carrying a gun before they can commit a criminal act. This detection can be achieved by the existing, and widespread, CCTV camera network in the UK. However, the performance of operators in interpreting CCTV imagery is variable as they are trying to detect essentially a very rare threat event. Additionally, current automated systems for detecting possible anomalous behaviour have been found to have varying success. We propose the development of a new machine learning system for the detection of individuals carrying guns which will combine both human and machine-based factors. Using selected CCTV footage which depicts people carrying concealed guns, and other control individuals, the proposal will establish what overt and covert cues (essentially conscious and subconscious cues) experienced CCTV operators actually attend to when identifying potential gun-carrying individuals from such CCTV imagery. In parallel, a machine learning approach will establish the machine recognised cues for such individuals. The separate human and machine cues will then be combined to form a new machine learning approach which will be fully tested. The system will be capable of learning and reacting to local gun crime factors which will aid its usefulness and deployment capability.