=Paper=
{{Paper
|id=Vol-3634/paper11
|storemode=property
|title=Organizing Transdisciplinary Research and
Innovation: the Case of the Socially-Aware Artificial
Intelligence Focus Area
|pdfUrl=https://ceur-ws.org/Vol-3634/paper11.pdf
|volume=Vol-3634
|authors=Loïs Vanhée
|dblpUrl=https://dblp.org/rec/conf/multittrust/Vanhee23
}}
==Organizing Transdisciplinary Research and
Innovation: the Case of the Socially-Aware Artificial
Intelligence Focus Area==
https://ceur-ws.org/Vol-3634/paper11.pdf
Organizing Transdisciplinary Research and
Innovation: the Case of the Socially-Aware Artificial
Intelligence Focus Area
Loïs Vanhée1,*
1
TAIGA Umeå’s Center for Transdisciplinary Artificial Intelligence for the Good of All and Department of Computing
Science, MIT-Huset, 901 87 Umeå
Abstract
Research, education, and innovation on trust in human-AI teams inherently involves inter/transdisci-
plinary considerations, which subsequently raises a wide array of challenges on how to produce such
research, from networks, funding, to alternate processes for producing science. Whereas the interest
around this topic mostly revolves around scientific productions, methods for organizing the required
underlying scientific productive system remain limited, from reaching unknowingly relevant researchers
to helping interested researchers to unfold, grow, and strive. This paper is dedicated to introducing one
of such academic environments through introducing the Socially-Aware Artificial Intelligence focus
area of TAIGA, Umeå’s center for Transdisciplinary Artificial Intelligence, as a research, education, and
innovation environment dedicated to support transdisciplinary AI research and how such an environment
can be deployed for serving the development of research on trust within human-artificial intelligence
teams in particular.
Keywords
transdisciplinarity, interdisciplinary AI, research organization, socially-aware artificial intelligence
1. Introduction
Creating trust within human-Artificial Intelligence (AI) teams inherently involves crossing a
multiplicity of frames, may they arise from different disciplines (e.g. psychology, computer
science, sociology, organizational theory, management), or from different sectors (e.g. academia,
research institutes, end-users, impacted businesses). This crossing of frames can take various
forms such as multidisciplinarity (i.e. applications of methods of different disciplines on the
same object of study); interdisciplinarity (i.e. developing new research topics and methods
on a given object of study at the interstice of multiple disciplines); and transdisciplinarity (i.e.
developing new knowledge about an object of study that combines perspectives of academic
and non-academic stakeholders) [1].
Despite the criticality of multi/inter/transdisciplinary research activities for studying trust in
Human AI teams, such research is, to a vast extent, carried through and forced to be carried
MultiTTrust: 2nd Workshop on Multidisciplinary Perspectives on Human-AI Team, Dec 04, 2023, Gothenburg, Sweden
*
Corresponding author.
$ lois.vanhee@umu.se (L. Vanhée)
https://www.umu.se/en/staff/lois-vanhee/ (L. Vanhée)
0000-0002-4147-4558 (L. Vanhée)
© 2023 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
Workshop
Proceedings
http://ceur-ws.org
ISSN 1613-0073
CEUR Workshop Proceedings (CEUR-WS.org)
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
through disciplinary lenses. From an organizational perspective, these research activities are,
for a vast majority, carried by classic scientific institutions, which are themselves structured
along disciplines from departments to communities, education, sources of funding, career
progression ladders, and recognition schemes (e.g. priority on journals or conference, author
ordering). This disciplinary mindset goes beyond social structures, being deeply internalized by
the researchers themselves, who, often unconsciously, carry practices, values, and assumptions
(on e.g. where to publish, what is quality science, down to the definition of what a well-formated
text is) arising from the primary disciplines they studied as undergraduates and their host
institutions. Incidentally, such disciplinary-centrism creates visible and invisible boundaries
[2] that inherently hinder the innovative mindset needed for transdisciplinary research and its
specific manifestation, such as studying trust factors in human-AI teams.
This paper is dedicated to providing an answer to the question: How can an existing classic
academic setting, organized along disciplinary faculties and departments, be adapted for best
enabling transdisciplinary research? Whereas this question is the object of study of entire
communities, this paper narrows the problem down to a case-study, i.e. a practical attempt
at organizing of research activities for enabling transdisciplinary research on Socially-Aware
Artificial Intelligence [3] (SAI), in particular through the studying the SAI focus area of Umeå’s
center for Transdisciplinary AI for the Good of All [4] (TAIGA). Specifically, this paper details
in Section 2 how the center and focus area are structured as to support transdisciplinary AI
research. Section 3 details how such organizational premises can be deployed for supporting the
initiation of new research track, in particular on how Umeå University research and researchers
can be bound to the MULTITTRUST interests and communities, crossing scientific tracks,
scientists, and structural prospects.
2. Organizing Transdisciplinary Research Through Centers and
Focus Areas
The SAI focus area within TAIGA is organized as to answer to the limitations raised by classic
disciplinary research organization. Briefly, TAIGA is dedicated to enable and develop transdisci-
plinary AI research, education, innovation, and social impact at the university level. TAIGA
is established over two framing principles. First, AI is framed as a transversal object of study.
Unlike classic perspectives that narrow AI down to technical (computer-science) perspectives,
TAIGA seeks to provide equal opportunities for all the relevant disciplines to partake in TAIGA’s
activities with an equal foot, including disciplines that can benefit from using AI methods in
their research and praxis (e.g. education, health) [5, 6] and disciplines that can study the ramifi-
cations tied to AI (e.g. how people relate to AI systems, AI as sociopolitical object [7]). Second,
framing AI research not only as an intellectual endeavour but also as an object of high salience
for society for which we are bound to provide answers.
As an organizational structure, TAIGA is organized around a coordination core and eight
focus areas that arise from eight different disciplines, titled: “AI in health and medicine”; “AI and
art”; “education and AI”; “understanding and explaining AI”; “critical, ethical, legal, social AI”;
“AI management”; “embodied interactive AI”; and “Socially-Aware Artificial intelligence” (SAI).
This organization is set as to enable the best connectivity across all the relevant disciplines and
Figure 1: Taxonomy of disciplinary, multidisciplinary, interdisciplinary, and transdisciplinary research.
Source: [1] (reprinted by creative commons permission)
the university.
The SAI focus area is primarily hosted by the computing sciences department. The SAI
focus area the most centred around the study of new AI methods that allow accounting for
and adapting to human social behavior (i.e. human-centered SAI) and/or accurately simulating
and replicating human social behavior (human-like SAI). Beyond technical and applicative
interests, the SAI focus area is dedicated to raising operational capacity at the university level by
providing state-of-the-art AI-powered competencies and research methods to interested other
departments (e.g. psychology, pedagogy, medicine, arts) as well as for fostering a university-wide
discussion of the ramifications of such technologies on society (e.g. sociology, organizational
science, business, law). The SAI focus area is also involved in reaching out to sectors and
actors beyond the university (e.g. businesses, region, other centers). As a general strategy for
achieving this mission, the SAI focus area is dedicated to identifying, enabling, and supporting
the development of all the prospective scientific niches available to Umeå University, crossing
all relevant pre-existing available resources (e.g. staff, skills, networks) and seeking to cover
and synergise all the university’s missions (i.e. research, education, innovation, social impact,
communication). This action is articulated along six activities, including fostering internal
collaborations through opportunities for academics to reach each other (A1), supporting their
academic missions (i.e. research (A2), education (A3), grant-writing and innovation projects
(A4)) tied to SAI, and reaching out to wider communities both within academia (A5) and beyond
(A6).
A1: Stimulating SAI Networks Growth through internal events (e.g. seminars, workshops,
pitch events, coffee gatherings); internal diffusion of available SAI competences and methods
(e.g. connecting computer scientists specialized in Natural Language Processing to pedagogy
and psychology); and workshops for identifying and spreading interests related to SAI (e.g.
regulation of SAI) in visits tailored to the interests of the disciplinary departments of the
university.
A2: Supporting Local SAI Research through identifying and organizing the internal
activities carried out by the university. This activity is structured along two axes. First, an
achievement-centered approach, through a research-focused bibliometric analysis of all Umeå
University’s SAI activities (e.g. listing all retrievable research papers, listing of research trends,
trends over time, productivity over time, key researchers of the topic and their contributions and
interests) and making this information available to other researchers [3]; second an ambition-
centered approach involving a systematic collection of current and future research interests
through forms and networking events.
A3: Development of Specialized Educational Activities through stimulating the creation
of transdisciplinary networks for enabling sufficient skill capacity on the teacher’s side and
interest on the student side as well as for acquiring mastery and legitimacy at the university
level for enabling change to occur administratively, as well as demonstrations of feasibility
through practical activities such as pedagogic hackathons on AI for social good1 .
A4: Supporting Funding and Development of new research activities through the dis-
semination of relevant sources of funding to the interested community, the organization of
pitch and case-specific or problem-specific events (e.g. AI for sustainability, AI for local urban
planning), and offering starting grants (seed funding) for providing the time to engage in the
preliminary research required for larger grant-writing.
A5: Academic outreach, both nationally and internationally through the organization
of international workshops on SAI topics (e.g. Interdisciplinary Design of Emotion-sensitive
Agents at the Autonomous Agents and Multi Agent Systems conference [8]); international
special interest groups on SAI-related topics (e.g. AI for crisis response [9]); special tracks
on SAI conferences (e.g. human-like deliberation and deliberation during crises at the Social
Simulation conference [10, 11]);
A6: Public Outreach through the organization and participation of seminars open to the
general public (e.g. the frAIday seminar series [12], a weekly seminar that regularly reaches more
1
https://www.umu.se/en/research/projects/autograide---automated-grading-of-ai/
than a hundred of participants from the university and beyond); organization and participation
in events involving innovation and society actors (e.g. dedicated encounters with stakeholders,
contribution to TAIGA conferences).
As an overall modus operandi, the resources of the SAI focus area are dedicated to cater for
qualitative spaces for long-term collaboration rather than for short-term reporting (e.g. number
of papers per year), which is expressed in at the implementation of the SAI focus area. For
example, the yearly seed funding call, dedicated to offer a small seminal budget (funding about
a month of research time plus a conference and/or technical resources) is organized as to best
support open-minded, creative, high-risk high-rewards collaborations either internal to the
university or external, as long as the university is involved. Within this frame, the format of
the application is purposefully constrained to be short (600 words), focusing on how the project
can develop SAI potentials for the university rather than on (e.g. budgeting) details, while
keeping relatively low demands in terms of reporting: the funding is serving the initiation of
collaborations rather than the other way around.
3. Case Study: Enabling Trust and Human-AI Teams Research
The aforementioned activities of the Social AI focus area allow for an actor external to Umeå
University (a research institution, a research group, an independent research) to effectively
identify prospects for collaborations with Umeå research actors in general. This section is
dedicated to specifying this identification for best enabling internal and external collaboration
on trust factors in human-AI teams research. The carried activities allow covering manyfactorial
considerations, including assessing the relevance of the research carried by the institution,
whether and how researchers (and subsequently, personal time, skills, interests, resources)
can be mobilized in new collaborations, and practical concerns for achieving collaboration in
practice (e.g. access to basic resources for initiating collaborations).
From a research perspective, a review of the SAI scientific activities brings into light three
lines of research carried out by Umeå University actors that touch upon trust in human-AI teams
research. A first line of research, directly related to the topic, bring forward interdisciplinary
studies of the impact of cultural factors in trust in human-AI systems: how culture impacts the
psychological processes involved in trust building and what does it entail when developing
human-AI systems [13, 14]. A second line of research captures human-AI teaming through the
lenses of adjustable autonomy and covers psychological factors, such as accounting for the
cognitive availability and demands made to involved humans [15], the integration of norms the
system should best try to comply to when interacting with humans [16], the development of in-
tuitive man-machine interaction media in which humans can advise the system on (un)desirable
courses of actions [17, 18] and more general frameworks on man-robot teams [19]. A third
line of research considers broader frames, bringing forward trustworthy AI considerations and
in particular on the topic of transparency and explainability [20, 21, 22]. From a researchers’
perspective, an array of profiles stand out. Two profiles are engaged in the framing of adjustable
autonomy along technical lenses with some involvement on the trustworthy AI track from
a technical AI perspective; a set of profiles focusing on the trustworthy/explainable AI track,
either from a technical perspective or a sociological perspective; and one profile engaged with
psychological and interactional factors in trust in human-AI teaming (culture, cognitive avail-
ability, anxiety) with a more interdisciplinary approach. Most of the involved profiles have
worked with both body-less AI and embodied (robotic) systems.
From a resource perspective, the analysis brings forward the availability of an array of
AI methods that can be relevant to the topic (e.g. robots, neural networks, natural language
processing), albeit most remain to be further articulated in the context of trust in human-
AI teaming. This analysis, based on the systematic mapping of already developed research
lines and involved researchers, acts as an effective approach for helping internal or external
actor to consider prospective collaborations along the line of trust in human-AI teams with
internal researchers, by bringing into light an array of feasible research directions and interested
collaborators. The next step consists in exploiting other means for engaging with interested
teams, either through direct contact with relevant researchers, or through engaging in SAI focus
area’s activities, such as seminars and networking events, or engaging with the SAI focus area
coordinator.
4. Conclusion and discussion
This paper brings into light an approach to research organization dedicated to creating and
fostering transdisciplinary research, innovation, education, and social impact at the level of
a university on trust in human-AI teaming, by describing one of such structure, namely the
Socially-Aware AI (SAI) focus area from TAIGA, the center for transdisciplinary artificial
intelligence of Umeå University. Through presenting the approaches and missions of the SAI
focus area and specifying them to the case of trust in human-AI teaming, the paper also shows
how such an organizational approach can operate as an effective support for connecting new
research ideas, internal and external researchers, and means for such collaborations to be
initiated (e.g. seed funding).
The SAI focus area is dedicated to providing the structures for alleviating the key pitfalls of
classic, discipline-centered organization of most universities by taking a transversal university-
wide stance on SAI research, education, and innovation: SAI methods are not (only) computer
science methods but an object of study of a wide array of disciplines. By its systematic identifi-
cation of former and ongoing research results and active researchers and through its proactive
approach to initiating and supporting collaborations, the SAI focus area enables for new research
tracks and collaborations to be added organically to existing objects of study already mastered
by (a part of) the university. This approach, which dis-encloses objects of study from specific
disciplines, allows for other disciplines to engage in the scientific debate surrounding these
objects of study as well as for scientists in seeking help from other disciplines to be provided
with the right contacts for this help, if reachable, to be offered (e.g. computer scientists seeking
supports from psychologists for validating their models).
While the SAI focus area remains too young for the potentials it creates to have yet matured in
finished research and funded projects (a process commonly scaling in years for interdisciplinary
research), early interventions have been positively received by a significant portion of the local
SAI community and by reached SAI researchers. Early contacts between computer scientists,
sociologists, philosophers, and law in workshops and hackathons facilitated by the SAI focus area
have already demonstrated the potential transformative approach on the research processes (e.g.
how models are built, assumptions), research outputs (e.g. qualities of the produced models) and
even research purposes (e.g. questioning the fitness of the models for society). As to exemplify
the case of trust in human-AI teams, the SAI approach put in motion in this paper brought into
light the potentials for collaboration along three research lines, from highly interdisciplinary
psychology-grounded perspective on trust, to technical perspectives on trustworthiness and
adjustable autonomy as well as fast tracks for these lines to be turned into working collaborations.
The options being laid out, now is the window of opportunity for internal and external actors
from all disciplines to step in enable the scaling up of this research.
5. Acknowledgement
The author acknowledges the support of TAIGA, Umeå’s center for transdisciplinary artificial
intelligence for having supported the production of this paper.
References
[1] G. Tress, B. Tress, G. Fry, Clarifying integrative research concepts in landscape ecology,
Landscape ecology 20 (2005) 479–493.
[2] F. Siedlok, P. Hibbert, The organization of interdisciplinary research: modes, drivers and
barriers, International Journal of Management Reviews 16 (2014) 194–210.
[3] L. Vanhée, Social ai research webpage, 2023. URL: https://sites.google.com/view/
taiga-socialai/research.
[4] U. University, Umeå university’s center for transdisciplinary for the good of all (taiga),
2023. URL: https://www.umu.se/centrum-for-transdisciplinar-ai/.
[5] K. Båth, E. Mårell-Olsson, J. Wedman, K. Danielsson, K. Grill, Humanities education in
the age of chatgpt: risks and opportunities, in: Higher Education in the Age of ChatGPT:
risks and opportunities, Online and in Umeå, Sweden, May 23, 2023, 2023.
[6] K. Kjell, P. Johnsson, S. Sikström, Freely generated word responses analyzed with artificial
intelligence predict self-reported symptoms of depression, anxiety, and worry, Frontiers
in Psychology 12 (2021) 602581.
[7] S. Lindgren, Critical theory of AI, John Wiley & Sons, 2023.
[8] L. Vanhée, Interdisciplinary design of emotion-aware agents (idea) workshop, 2023. URL:
https://en.uit.no/project/idea.
[9] F. G. C. Kammler, L. Vanhée, Building resillience with social simulations (bricss) special
interest group, 2023. URL: https://sites.google.com/view/bricss/home.
[10] E. S. S. Association, Simulating in crises special track at the social simulation conference
2023, 2023. URL: https://ssc23-sphsu.online/simulating-in-crises/.
[11] E. S. S. Association, Sense and sensibility special track at the social simulation conference
2023, 2023. URL: https://ssc23-sphsu.online/sense-sensibility/.
[12] U. University, Fraiday webpage, 2023. URL: https://www.umu.se/en/research/our-research/
features-and-news/artificial-intelligence/fraiday/.
[13] M. Borit, L. Vanhée, P. Olsen, Understanding the impact of culture on cognitive trust-
building processes: How to increase the social influence of virtual autonomous agents., in:
TRUST@ AAMAS, 2014, pp. 100–111.
[14] M. Borit, L. Vanhée, P. Olsen, Towards enhancing trustworthiness of socially interactive
and culture aware robots (2014).
[15] L. Vanhée, L. Jeanpierre, A.-I. Mouaddib, Optimizing requests for support in context-
restricted autonomy, in: 2021 IEEE/RSJ International Conference on Intelligent Robots
and Systems (IROS), IEEE, 2021, pp. 6434–6440.
[16] L. Vanhee, H. Aldewereld, F. Dignum, Implementing norms?, in: 2011 IEEE/WIC/ACM
International Conferences on Web Intelligence and Intelligent Agent Technology, volume 3,
IEEE, 2011, pp. 13–16.
[17] L. Vanhée, L. Jeanpierre, A.-I. Mouaddib, Augmenting markov decision processes with
advising, in: Proceedings of the AAAI Conference on Artificial Intelligence, volume 33,
2019, pp. 2531–2538.
[18] L. Vanhée, L. Jeanpierre, A.-I. Mouaddib, Augmenter les processus de décision via des
conseils, in: Journées Francophones sur la Planification, la Décision et l’Apprentissage
pour la conduite de systèmes (JFPDA), 2019.
[19] M. Chiou, S. Booth, B. Lacerda, A. Theodorou, S. Rothfuß, Variable autonomy for human-
robot teaming (vat), in: Companion of the 2023 ACM/IEEE International Conference on
Human-Robot Interaction, 2023, pp. 932–934.
[20] R. H. Wortham, A. Theodorou, Robot transparency, trust and utility, Connection Science
29 (2017) 242–248.
[21] D. Calvaresi, Y. Mualla, A. Najjar, S. Galland, M. Schumacher, Explainable multi-agent
systems through blockchain technology, in: Explainable, Transparent Autonomous Agents
and Multi-Agent Systems: First International Workshop, EXTRAAMAS 2019, Montreal,
QC, Canada, May 13–14, 2019, Revised Selected Papers 1, Springer, 2019, pp. 41–58.
[22] S. Knapič, A. Malhi, R. Saluja, K. Främling, Explainable artificial intelligence for human
decision support system in the medical domain, Machine Learning and Knowledge
Extraction 3 (2021) 740–770.