The field of digital humanities (DH) is born from the encounter between traditional humanities and computational methods. But the nature of this encounter, between two fields that until recently were considered inherently seperate, is still being determined: identifying the potential fruitful interactions can be established between humanities and computational sciences is an ongoing process of discovery, and the definition of DH is evolving with this process. “Digital humanities” is currently still a strategic term that covers most of the changes in topics and methodologies that are happening in humanities as a result of the boost given by the introduction of new information technologies. It is a term that will probably come to indicate a phase of transition, looking towards a future in which computational tools will be an integral part of society and culture, and also of social and cultural studies.
Of the many areas of computational sciences that can play and are playing a part in the evolution of DH, nowadays a key role is assumed by artificial intelligence (AI), which aims to develop computational agents able to perform tasks that are complex enough to be generally considered a prerogative of intelligent entities. In recent years, successful applications of AI have become more widespread, and AI has gained considerable attention outside the small community of practitioners in academia and industry, attracting media coverage and becoming part of the popular culture. AI methods are already present in many tools that are widely used in DH. A successful example of the use of AI techniques is in the area of information retrieval, where texts are stored in document management systems to help users quickly retrieve the information they are interested in1, but the future interplay between AI and humanities is potentially huge, going well beyond the use of automated tools to speed up and refine tasks that are seen as boring and repetitive. There is a real potential for interdisciplinary efforts to change both fields.
AI techniques can enable the exploration of territories that were previously inaccessible, such as in education sciences, where dedicated software and robots are being developed and promise to play an essential role in teaching and learning.2 Another potential are of application is law, where the use of knowledge representation and reasoning techniques allows the inference of new knowledge from legal data.3
One of the main challenges for AI at present is bridging the gap between different technical approaches: there are currently the sub-symbolic tools focused on autonomous or semi-autonomous learning, based on the use of neural networks and statistics, and also symbolic logic-based approaches for management of data and reasoning tasks. Integrating both approaches is the central issue for some very popular sub-areas, such as explainable AI (XAI). The AI tasks used in social sciences can help push the boundaries of research in this direction.
Within the DHH team, Shohreh Haddadan’s work4 is an example of an advanced form of interaction between social studies and AI. Her research area is argumentation mining; a pioneering topic in the field of natural language processing:5 given a text, the software will be able to individuate the distinct components of an argument – the main statements and any additional statements that support or are in conflict with them. Her software tools have been developed to analyse the argumentative structure of US presidential debates, with the aim of making an essential contribution to research questions like “How are political arguments structured?” and “How have argument structures developed over time and how do they vary across different topics?”. Haddadan considers a form of analysis that is a prerogative of historians and philosophers – namely, how contemporary political argumentation works, and proposes to incorporate into its methodology formal tools from the area of formal argumentation,6 a line of research that is part of AI; moreover, her software tools allow this tiype of analysis to be automated. But the contribution made by her research is not only unidirectional, from AI to historical studies: The analysis of political debates is not simply a use case scenario for the argument mining software; in our view this line of research can be taken further, resulting in the introduction of reasoning tools that can be used to identify particular patterns in the political debates, such as argumentative fallacies. In other words, research questions that are typical of humanities allow us to push the boundaries of AI, looking for the development of architectures incorporating AI research approaches that had yet to be considered; in this case natural language processing and automated reasoning.
Haddadan’s work is developed in connection of the Individual and Collective Reasoning (ICR) group, which is part of the DHH Unit and of which also the authors are members. ICR is a research team dedicated to AI research; although its members and collaborators work in computer science, many of them have multi-disciplinary backgrounds, ranging from mathematics, philosophy, economics, law, and linguistics, and the group has experts in all the main technical approaches to AI: symbolic (knowledge representation and reasoning), sub-symbolic (machine learning), and situated cognition (robotics). In the future the group plans to increase its efforts in DH research in two areas in particular: ethics and art.
AI Ethics
AI introduces new artificial actors into the social landscape, such as social robots7, automatic decision-making systems, and self-driving cars8. Various individual disciplines including philosophy, law, sociology, economics, and AI have created spaces for discussing the ethical, legal, social and technological aspects of the use of AI systems, but until now the discussions have generally remained confined to each seperate discipline.
In recent years possible interplays of computational tools and philosophical research have started to be investigated, especially by Prof. Christoph Benzmüller, who until recently was a visiting researcher in the ICR group. Using automated reasoning tools to analyse complex philosophical arguments, it was possible to check the logic of the arguments presented in complex essays or even to develop new results that would have been virtually impossible to achieve without the assistance of appropriate automated reasoners.9
What we see as the nextstep in the discussion of ethical problems, in particular when connected to AI, is the definition of a solid ground for collaboration among the fields of formal ethics, AI and law, and AI ethics. The fields of formal ethics and legal reasoning can bring new tools that can be applied in the field of AI ethics, such as deontic logic, decision theory, social choice theory and formal argumentations, while in return, the field of AI ethics can offer case studies and insights into specific problems. The development of AI ethics for a system in which humans and natural agents interact requires the redefinition of fundamental notions,10 such as agency, individuality, responsibility (especially considering artificial actors and complex multi-agents systems) and fairness, as well as new notions that have become more relevant with the emergence of new AI technologies, such as biased information: avoiding the latter when training artificial (and natural) actors has become an important ethical issue.
In our view a keychallenge is the development of an effective system that facilitates the formalisation of ethical theories and can assist agents in evaluating and taking decisions in specific situations.
AI & Art
New technologies change humanity. They invariably affect economies and therefore social structures, but in the case of AI they also directly influence some of the basic notions associated with what it means to be human, as mentioned above: notions like individuality, intelligence, agency, responsibility, society and so on. The redefinition of some of these notions is intricately linked with the AI endeavour: the same test that was initially proposed to evaluate AI, the famous Turing Test, reminds us that the goal of AI, at least in its initial phase, was to break down the distinction between artificial and natural agency, making the two indistinguishable from the point of view of social interaction.
This need to reinterpret basic concepts and relationships has influenced many disciplines since the 1980s, in particular philosophy, and it affects also other methods for interpreting reality, like art.
The ICR group is organising and coordinating an AI&ART Pavilion that will be part of the series of events for ESCH2022 – European Capital of Culture.
New technologies affect the way in which we create art, enabling the creation of new media and new forms of interaction between the public and the artwork. Furthermore, these art productions can help us investigate and reinterpret traditional notions that may be seen as being jeopardised by AI.
Conclusion
Humanities offer a means of analysis, a way of interpretating what it means to be human and to be part of society. The interaction between humanities and AI is unavoidable. Moreover, while the impact of AI on society is becoming a main topic of investigation for many disciplines, the interplay between AI and humanities goes beyond the former being a potential object of investigation of the latter. There is a stronger potential interplay at a theoretical level, since many notions that are investigated in social sciences are put under considerable pressure by new AI research and therefore need to be redefined in a way that takes new computational models into consideration. Moreover, new computational tools also offer new formal languages and reasoning power that can influence the methodologies of some disciplines, while AI itself is enriched by modelling scenarios that can guide future developments. In this paper we have looked at some possible ways in which AI and humanities can interact, thereby helping us to review and develop our interpretation of the world.
Edited by Juliane Tatarinov; English review by Sarah Cooper
Notes
- P. Casanovas, M. Palmirani, S. van Engerse, T. Peronid, and F. Vitali, “Special Issue on the Semantic Web for the Legal Domain Guest Editors’ Editorial: The Next Step,” Semantic WebJournal, vol. 7, no. 3, pp. 213-227, 2016.
- https://www.fnr.lu/research-with-impact-fnr-highlight/from-lab-to-startup-luxai/
- https://www.mirelproject.eu/
- S. Haddadan, E. Cabrio, S. Villata, “Yes, we can! Mining Arguments in 50 Years of US Presidential Campaign Debates”, Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL 2019), pp. 4684–4690, 2019.
- M. Stede and J. Schneider, “Argumentation Mining”, Synthesis Lectures on Human Language Technologies, vol. 11, no. 2, pp. 1–191, 2018.
- P. Baroni, D. Gabbay, M. Giacomin, L. van der Torre (eds.), “Handbook of formal Argumentation”, College Publications, 2018.
- R. Bemelmans, G. J. Gelderblom, P. Jonker, L. De Witte, “Socially assistive robots in elderly care: A systematic review into effects and effectiveness”, in Journal of the American MedicalDirectors Association, Vol. 13, no. 2, pp. 114-120, 2012.
- A. Shariff, J. F. Bonnefon, I. Rahwan, “Psychological roadblocks to the adoption of selfdrivingVehicles”, in Nature Human Behaviour, Vol. 1, no. 10, p. 694, 2017.
- D. Kirchner, C. Benzmüller, E. N. Zalta, “Computer Science and Metaphysics: A Cross-Fertilization”, in Open Philosophy, Vol. 1, no. 2, 2019; D. Fuenmayor, C. Benzmüller, “A Computational-Hermeneutic Approach for Conceptual Explicitation”, https://arxiv.org/abs/1906.06582, 2019.
- D. Dennet, “The Intentional Stance” (6th printing), The MIT Press, 1996; M. E. Bratman, “Shared Agency: A Planning Theory of Acting Together”, Oxford University Press, 2014.