About the sandpit
We will shortly invite applications from individual researchers from the UK, Canada and US to take part in a collaborative, interdisciplinary funding sandpit.
A sandpit is a facilitated workshop process through which new project teams and research propositions will be co-developed and funded.
Participants apply as individuals, and applications are generated during the workshops, not in advance.
This sandpit is jointly delivered by the UK’s AHRC through the International Science Partnerships Fund (ISPF), and SSHRC in partnership with the Alan Turing Institute and The University of Edinburgh.
The sandpit is directed by Professor Drew Hemment, lead of the Doing AI Differently initiative, based at The Alan Turing Institute and The University of Edinburgh.
Sandpit process
There are two elements to the sandpit process:
- an expressions of interest stage where sandpit attendees will be decided
- the sandpit itself where attendees will form teams, develop project ideas and submit an application for research funding
The expressions of interest stage will include:
- submission of a short questionnaire
- review of questionnaire responses and sending of invitations to successful applicants
The sandpit process will include:
- a three-day in-person residential workshop in Canada (February 2026)
- a virtual workshop in April 2026
- director and mentor-facilitated application development, peer review and cohort support
- supporting up to four projects, which must begin on 1 October 2026
When the expressions of interest opportunity is live, you must submit an expression of interest as an individual, not a team or group.
We are looking for attendees without preconceived project ideas or teams.
The Doing AI Differently White Paper will be the primary reference document for this sandpit.
It sets out the research vision, theoretical foundations and strategic context.
We strongly encourage you to read it before submitting an expression of interest.
What is the thematic focus of the sandpit?
We are at a pivotal moment in AI design.
Contemporary AI systems increasingly produce cultural outputs, such as text, images and multimedia that resemble human cultural artefacts in unprecedented ways.
This represents AI’s ‘qualitative turn’: a fundamental shift toward systems whose inputs and outputs are deeply embedded in human cultural contexts.
This transformation creates an urgent need for humanities and arts perspectives to shape AI’s foundational design, training, and evaluation, not merely analyse its outputs.
The interpretive technologies theme addresses this need by supporting research that integrates humanistic methodologies directly into AI system design.
What are interpretive technologies?
Interpretive technologies refer to AI systems designed to engage with ambiguity, context-dependence, and plurality as core capabilities.
Rather than producing monolithic outputs, these systems can respond more appropriately to:
- diverse cultural contexts
- maintain sensitivity to meaning
- handle the semantic depth that characterises human interpretation
This work explores how humanistic insight can inform systems that do not attempt to imitate human judgement, but can respond contextually to the complexity of human and more-than-human experience while remaining computationally tractable.
We seek research that demonstrates mutual enrichment between technical and humanistic methods, where technical innovation both draws from and challenges interpretive frameworks.
Research we’re looking for
We invite novel, transdisciplinary research that advances knowledge in both humanities and arts and technical domains while contributing to shared research infrastructure.
Projects funded through the sandpit should demonstrate clear potential to:
- develop functioning prototypes that show measurable improvements in AI systems’ interpretive capabilities
- create new evaluation frameworks and benchmarks that assess cultural sensitivity, contextual awareness and perspectival reasoning
- integrate humanities or arts methodologies into technical design pipelines, showing direct influence on system design
- contribute to an open ecosystem of tools, methods and resources for interpretive AI
In scope
In scope examples include:
- novel training strategies or evaluation methods that enable AI systems to represent and process multiple cultural perspectives, including metrics that capture semantic depth and nuance
- contextual embedding approaches that allow systems to respond appropriately to specific historical, cultural or social contexts
- frameworks for expressing ambiguity, uncertainty, and plurality, including through interactive or multimodal outputs
- integration of hermeneutic, aesthetic, or narrative methodologies into system training or inference
- collaborative co-creation frameworks that use storytelling, speculative design, or artistic practices to shape how AI systems are developed and evaluated, enabling systems to reflect diverse human imaginaries rather than replicate them
- applied explorations in domains like sustainability or healthcare, where interpretive capabilities may enable context-aware, culturally responsive systems
Out of scope
Out of scope examples include:
- post-hoc ethical analysis of AI systems without influencing their fundamental design or training
- research that treats arts and humanities perspectives as supplementary or advisory, rather than central to the AI design process
- use of AI as a tool for humanities research without reciprocal influence on AI system capabilities
- approaches that do not demonstrate potential for substantive collaboration between humanities scholars and technical AI researchers
Collaborative requirements
All research must involve meaningful collaboration between humanities or arts scholars and technical AI researchers.
We particularly welcome researchers whose work opens new transdisciplinary territory beyond established fields, and approaches that engage with communities whose knowledge systems or lived experiences are often excluded from AI design.
All projects developed and funded via the sandpit must have at least one Canada-based and one UK-based lead.
Applications without both a UK and Canada lead will be rejected.
US researchers are eligible to be funded via the sandpit as project co-leads.
Expected outcomes
While we do not expect to fundamentally redesign AI systems in a single project, or group of projects, these initiatives will help lay the groundwork for longer-term innovation.
Funded research will produce:
- working demonstrations of interpretive AI capabilities with measurable improvements
- shared evaluation frameworks and benchmarks for assessing interpretive depth
- open-source tools and methodologies that contribute to an emerging research infrastructure
- documentation of transdisciplinary approaches that advance both technical and interpretive knowledge
- contribution to the global research community established through the Doing AI Differently initiative
The workshops themselves are expected to be a valuable and generative experience, with the potential to significantly advance thinking and collaboration in this emerging area.
Participants will be supported throughout by a team of expert mentors, who will work closely alongside Programme Director Drew Hemment to guide and shape the journey.
Up to four projects may be supported.
Staff costs for Canadian team members are not included in the grant amount as salary costs are covered elsewhere. According to Canada’s Tri-agency Guide on Financial Administration, SSHRC grant funds must not be used to pay compensation to applicant team members.
