Conference Agenda

If someone were to argue that Extensible Markup Language [XML] and Formal Ontologies [FOs] have little in common, they would have many strings to their bow. The main one, for my part, is this. XML is, as its name suggests, a markup language – or rather, a metalanguage that allows users to define their own customized markup languages (Attenborough 2003). FOs are neither languages nor metalanguages; they are artifacts specified by ontological languages (Gruber 2009). And XML is not even one of those languages. As for the “little” that XML and FOs have in common, there is one similarity that caught my attention. Both XML and FOs have something to do with partitioning. XML partitions data using elements. FOs partition domains of interest by means of representational primitives. Precisely from the FOs partitioning, the philosophical debate has outlined an epistemological view on FOs, namely perspectivism. For this kind of perspectivism – which does not coincide with perspectivism in the philosophy of science – partitioning a domain means making a mental division between those entities on which we focus and those that fall outside our (domain of) interest. According to this view, such a partitioning provides a perspective on the domain. Moreover, as perspectivism holds, whatever domain we consider, there can in principle be multiple, equally valid and overlapping perspectives on the same domain.

Now, in Tambassi (2023) it has been argued that perspectivism is not just one of the philosophical views that populate the debate on FOs, but an underlying assumption of FOs. In other words, FOs are perspectivist. In this talk I investigate whether the same is true of XML. I begin by defining FOs and presenting the main claims of perspectivism. The idea is not to show the perspectivism of FOs, but rather how these claims apply to FOs. This is also to avoid any overlap with Tambassi (2023). Then I move on to XML, showing both the perspectivism of XML and how the claims (of perspectivism) apply to XML. The argument is based on a parallelism between FOs and XML. More specifically, the facets of perspectivism on FOs that I present in the first part of this talk correspond to the facets of perspectivism on XML that I present in the second part. This is not intended to exhaust the ways in which perspectivism relate to FOs and XLM, nor the debate about FO and XML partitions. The only aim is to clarify whether and how XML is perspectivist. And on the grounds that XML and FOs have little in common, it is not even excluded the chance that perspectivism applies differently to FOs and XML.

The purpose is therefore purely speculative. I believe that discussing whether XML is perspectivist may help to clarify some of the theoretical assumptions of this markup metalanguage. More generally, the idea is that, since the creators of markup (meta)languages develop those languages under the guidance of some theoretical assumptions, for the sake of methodological accuracy those assumptions should be subjected to critical analysis rather than remain implicit and unexamined. The focus on XML is not accidental. First, XML is still widely used, and there are many other markup languages based on XML. This means that this critical analysis is, at least in principle, extendable to other markup languages. Second, XML not only supports the exchange of data, but it is also both human- and machine-readable. In other words, XML – like FOs – supports communication between humans, between humans and machines, and between machines (Goy and Magro 2015). And while supporting this communication is certainly not the prerogative of XML and FOs alone, we cannot even rule out the possibility that determining whether XML is perspectivist may also shed new light on some of the theoretical assumptions behind such communication.

References

Attenborough M (2003) Mathematics for Electrical Engineering and Computing. Newnes

Goy A, Magro D (2015) What are ontologies useful for? Encyclopedia of information science and technology (pp. 7456–7464). IGI Global

Gruber TR (2009) Ontology. In Liu L, Özsu MT (eds) Encyclopedia of Database Systems. Springer. DOI: https://doi.org/10.1007/978-0-387-39940-9_1318

Tambassi, T. (2023) On Perspectivism of Information System Ontologies. Foundations of Science, 2023, DOI: https://doi.org/10.1007/s10699-023-09900-5

In some remarks on calculation from the Remarks on the Foundations of Mathematics (RFM 1.149), Wittgenstein describes a thought experiment where a group of people have the odd practice of valuing wood by surface area rather than volume. This provides a particularly vexing example of a logically alien thought: calculating and trading which does not map onto the norms of reason we hold. Nevertheless, we hesitate to dismiss the practice as irrational; surely, there must be reasons since we recognize the agents as human beings and, thus, rational beings. Indeed, this is the suggestion that Wittgenstein leaves us with at the end of the remark.

This thought experiment poses the question: how should we apply the principle of charity in cases where we encounter strange cases of rule-following? When someone is performing actions or making claims that purport to satisfy some norm, but one which fails to make sense according to our norms of reason, how should we interpret it? These questions were at the heart of the “rationality debates” in the philosophy of social science, the discussions in social science about how we should, or even whether we can, find seemingly irrational beliefs intelligible (e.g., Wilson 1970, Hollis and Lukes 1982, Risjord 1993; 2001). This question turns on which norms are guiding in our assessment of prima facie unintelligible behaviors, specifically when does it make sense to say that such a system follows a rule that is intelligible to us.

The problem has received new relevance in the age of chatbots. The current large language models (LLMs), such as ChatGPT and Gemini, often make claims that appear to satisfy norms—they make claims that are truth-evaluable and appear to be relevant and useful. Moreover, they are correct more often than not in a number of domains. But, at the same time, the claims are often not just false but absurd, as when ChatGPT suggests that a good way to get a couch on the roof is to lower it from a truck bed.

In this paper, we argue that revisiting the rationality debates can help address how to make sense of these systems. Specifically, revisiting debates about the principle of charity for interpreting alien others highlights the tensions with the different ways we might respond to these models. We focus on two different approaches along with Risjord (1993,2001): first, we present the ‘rationalist first’ approach—those who argue that in the first instance we should adopt our own norms of rationality as standards for evaluating these models—that, when we treat the model as having true beliefs and being rational, we are assuming they are participating in a human form of life, one where their practices are broadly commensurable with our own. But, we show, this approach faces insurmountable problems, regardless of whether it treats LLMs as rational or irrational. Second, we follow the ‘explanatory coherence’ approach—those who argue that our goal should be to maximize the explanatory power of our account while minimizing error and inconsistencies. We argue the latter approach is better for understanding LLMs. The upshot is that the rationality debates throw light on different approaches we can take towards these models, but also highlight how logically alien these machines are.

Selected Bibliography:

Hollis, Martin and Steven Lukes, eds. (1982). Rationality and Relativism. Cambridge: MIT Press

Risjord, M. W. (2000). Woodcutters and witchcraft: Rationality and interpretive change in the social sciences. Suny Press.

Risjord, M. (1993). “Wittgenstein's Wooductters: The Problem of Apparent Irrationality.” American Philosophical Quarterly, 30(3), 247-258.

Wilson, Bryan, ed. (1970). Rationality. Oxford: Blackwell.

Wittgenstein, Ludwig (1967) Remarks on the Foundations of Mathematics Translated by G. E. M. Anscombe, Cambridge: MIT Press: p. 44e.

Bocong Li (2021) argued that philosophy of engineering, being a philosophy of action, implies a process-centered vocabulary, rather than the object-centered language of epistemology of science. Philosophy of engineering is a “verbal concept system”, focusing on the action and event more than on the objects and properties, while epistemology is a “nominal concept system”. Since engineering practice relies on (although cannot be reduced to) application of engineering science, how do the languages of engineering science and practice translate into each other?

Philosophy of science differentiates between the concepts and representations of time in different disciplines (e.g., the physical, geological, biological time) and types of systems (open, closed, living, technical systems). In engineering, the universalisations and abstractions are the tools for the concrete applications, which are contextualized in time and space (Banse&Grunwald, 2009). Engineering projects are integrating the technical, organizational, political, economic timeframes and the temporal dimension of the human lifeworld for workers, users and the engineers themselves.

STS studies have shown that engineering work and knowledge operate at the intersections of timeframes and temporalities (Shih, 2009; Gainsburg et al., 2010, Vinck, 2011). The complexity of engineering practice implies coordination between the quantification of time in the natural and engineering sciences in modeling and experimentation, the control over time which is central to the modern management, standardization and planning, and the collective and individual temporality of action.

The synchronization and parallelism of these different dimensions requires a process of translation. Such translation occurs in construction of the networks and objects in engineering (Suchmann, 2000; Bucciarelli, 2002; Balco, 2003). This report focuses on the way the different languages of engineering (formal, natural, and visual) coordinate the processes, events and actions. The notions that are used to describe the technical and the social process alike (such as stage, phase, iteration, progress, effect, contingency, etc.) are of special interest from the point of view of pragmatics of engineering language. The interviews with engineering practitioners and document analysis of engineering projects reveal the contextual variety of the formal and colloquial notions of time and its associates. In turn, such narratives organize and give meaning to engineering practice as it evolves in different temporal perspectives.