Source-linked AI summary
Reciprocal Collaboration: how lessons from convergence in GLAMs can enhance interdisciplinary AI research
Amber L. Cushing, Suzanne Little, Giulia Osti
TL;DR
Interdisciplinary AI collaborations can struggle to translate concepts and values across disciplines, while some partnerships leave domain researchers with limited influence. The paper draws on GLAM convergence to propose a reciprocal, reflexive framework of five practices for collaboration.
Problem
Interdisciplinary AI collaborations can involve unidirectional communication, unequal partnerships, and risks of overlooking domain methodologies and expertise.
Source: §2.2 · pp. 4–6Method
The paper draws lessons from GLAM convergence to develop a reciprocal approach grounded in reflexivity and written shared values.
Source: §3.1 · p. 6, §3.3 · p. 9, §4 · p. 10Results
The authors propose five practices intended to improve communication and understanding between disciplines and identify substantial impacts and improvements.
Source: §5 · p. 15Takeaways & Limitations
GLAMs may serve as a testbed for refining interdisciplinary AI integration before the approach is applied in other scientific contexts.
Source: §4 · p. 11
Abstract
from arXiv · showhide
The need for collaboration between diverse fields of research is increasingly recognised as important by research funding agencies. A significant driver of this need is the current revolution in artificial intelligence (AI) and related technologies. There is a growing interest in the potential impact of AI in different fields including the methodologies they use and the resulting advances in new knowledge, new access and enhanced productivity. However, there is also a corresponding increase in concern about the fundamentals of AI technologies and the way in which trans and/or interdisciplinary research is approached. The resulting collaboration too often ends up as a one-way street where the domain partner acts only as an information provider. For example, the contribution of the AHSS partner might be limited to providing insight about ethics and/or the technology partner may only provide a service to build applied AI-based solutions. In response to this problem, we propose a reciprocal approach to collaboration where both partners seek to understand, cooperate and identify jointly significant impacts. In this paper we explore this relationship between cultural heritage institutions (GLAMs), Arts, Humanities & Social Sciences (AHSS) research and technology-led AI research, especially the impact of current technological advances in AI. Drawing from the history of convergence in GLAM studies, we propose five key practices to form a framework for greater understanding across this divide.
2 Strains of interdisciplinary AI research collaboration
Research on interdisciplinary AI collaboration highlights both the potential contributions of AHSS and persistent difficulties translating concepts and values across disciplinary boundaries. The literature points to unequal partnerships, disciplinary tensions, and risks when AI work overlooks domain expertise.
2.1 Key examples
- AHSS contributions to AI include archival approaches to dataset documentation and humanities and social-science insights about power, culture, complexity, and social impacts.Datasheets-inspired work in digital cultural heritage seeks both human readability and machine readability; “Interpretive AI” is described as handling plurality, ambiguity, and contextual meaning.Source: §2.1 · pp. 3–4
- Workshops introducing humanities-based intersectional methods to computing students found openness alongside tension with results-oriented disciplinary conventions.Research on critical fabulation in participatory design also describes reflexivity as a way for researchers to reflect on their positionality.Source: §2.1 · p. 4
2.2 Existing difficulties with interdisciplinary AI research
- The “parachute approach” brings outsiders in to diagnose problems but excludes local expertise and limits who can shape solutions.In interdisciplinary projects, this can mean consulting an outside specialist without inviting them to influence the overall project.Source: §2.2 · p. 5
- AHSS scholars report co-opted concepts, limited leadership, mismatched priorities and incentives, and few opportunities for equal partnerships in AI research.Reported challenges also include little time to integrate perspectives, limited institutional support, and unwillingness to engage with humanities beyond philosophy.Source: §2.2 · p. 5
- AI development can overlook domain methodologies, amplify mistrust and bias, and exploit data or knowledge without delivering true benefits.The paper frames these risks alongside claims that AI can bring progress, efficiency, and societal or economic impact.Source: §2.2 · pp. 5–6
3 Insights from the history of convergence in GLAM Studies
GLAM convergence offers lessons in coordinating institutions, translating critical concepts into practice, and adopting technology in light of shared values. Its outcomes are not uniformly positive, making structures for communication and recognition of expertise important.
3.1 A brief introduction to convergence in GLAMs
- GLAM convergence merges cultural heritage institutions or collections and can involve shared standards, organizational arrangements, research labs, or educational programs.Its development has been intertwined with technology intended to enhance access, but some convergences were later disbanded after policy changes.Source: §3.1 · p. 6
- Converged GLAM institutions have reported stronger communication, mutual respect, and collaboration, but convergence alone does not guarantee these outcomes.Clear responsibilities and recognition of specialist expertise matter; convergence treated as a bureaucratic exercise or a way to cut roles can threaten professional identity and leave staff dissatisfied.Source: §3.1 · pp. 6–7
3.2 Translating critical work to applied practice via reflexivity
- GLAM studies translates critical concepts into applied practice, with reflexivity supporting examination of researchers’ positions, power, assumptions, and research processes.Examples include applying ethics of care to archival practice and sensemaking theory to reference interviews; education can also teach practitioners how to apply theory.Source: §3.2 · pp. 7–8
3.3 Technology adoption in the context of convergence
- GLAM technology adoption is shaped by shared values and professional standards, not only technical considerations.Digital archivists have considered how AI may challenge professional expertise, while decisions to adopt technology can prompt review of institutional missions and best practices.Source: §3.3 · p. 9
- GLAM studies offers a potential roadmap for interdisciplinary AI collaboration, where existing arguments lack guidance for developing integrated concepts, hypotheses, and research strategies.The paper connects this gap to GLAM experience aligning applied practice with jointly developed standards and best practices.Source: §3.3 · p. 9
4 A reciprocal approach to guide interdisciplinary collaboration in AI research
The paper proposes reciprocal collaboration grounded in reflexivity and shared values, drawing on GLAM convergence to address barriers to interdisciplinary AI research. Its practices include clarifying language and research problems, examining values, and making individual habits and beliefs explicit.
4 A reciprocal approach to guide interdisciplinary collaboration in AI research
- A reciprocal approach is an ongoing, co-constituted collaboration in which written shared values guide decisions and mutual reflection surfaces differing voices and positions.The approach treats collaboration as capable of shaping project findings and rejects “parachute” work that limits bilateral decision-making and ignores disciplinary norms.Source: §4 · p. 10
- GLAMs are proposed as a testbed for refining integration because they have experience translating concepts into practice and developing shared standards for technology adoption.The authors suggest this setting may allow the process to be refined with less friction before application in other scientific contexts.Source: §4 · p. 11
4.1 Practice being open and upfront about jargon and disciplinary terms
- Researchers should clarify disciplinary jargon upfront because common terms, such as “tool,” can carry different meanings across fields.The paper contrasts social-science uses of “tool” as a method, instrument, approach, or framework with computing uses as a digital application.Source: §4.1 · p. 11
- Making space to ask about terminology and disciplinary tensions can support jointly created best practices for AI development.The paper presents this as reflexive practice that may help teams implement AHSS concepts and theories in machine-learning and computer-vision work.Source: §4.1 · p. 12
- Collaborators’ disciplinary lenses can shape what they consider novel, impactful, or successful, including preferred research pace and publication venues.The paper describes tension between slower attention to social implications and faster pursuit of pragmatic applicability.Source: §4.1 · p. 12
- Reflexivity treats tensions over research standards as prompts for self-reflection, rather than issues that must be eliminated.A conflict between humanities-based work and expectations of model performance can prompt researchers to reconsider what counts as valid and valuable research.Source: §4.1 · p. 12
4.3 Develop a common understanding of the most pressing research problems in relation to AI
- Researchers should make their understandings of AI research problems explicit and revisit them through discussion to identify areas of overlap.AHSS literature often frames fairness, bias, ethics, and accountability as societal problems, while computer-vision and machine-learning work more often frames them as problems of program behavior and performance.Source: §4.3 · pp. 13–14
- AI evaluation often relies on accuracy and other narrow metrics, while interdisciplinary work can examine usefulness in relation to professional missions and users’ lived experiences.The paper argues that teams should identify how values overlap and gather data about user experiences to develop a holistic view of technology adoption.Source: §4.3 · p. 14
5 Conclusion: A call to arms
The paper proposes five practices for reciprocal collaboration between AHSS and AI technology researchers, drawing on GLAM convergence to improve understanding and identify substantial impacts. It also proposes using the practices in early-career researcher training.
5 Conclusion: A call to arms
- Five practices, guided by GLAM convergence, are proposed to support reciprocal collaboration between AHSS and AI technology researchers.The framework is intended to assist project development and execution.Source: §5 · p. 15
- The practices aim to enhance cross-disciplinary communication and understanding and help identify substantial impacts and improvements.Source: §5 · p. 15
- Training early-career researchers is another proposed use, including translating jargon, making approaches explicit, and acknowledging different understandings of outputs and impacts.The paper connects this aim to EU doctoral education objectives.Source: §5 · p. 15
- Future work will use reflexive and autoethnographic practices in PhD training to operationalise the concepts and refine their use in cutting-edge research.Source: §5 · p. 15