Introduction

The current (still limited) use of the notion of informativeness in the domain of information system ontologies [ISOs] (§ 2) seems to indicate that ISOs are informative if and only if they are understandable for their final recipients (§ 3). This paper discusses some theoretical issues emerging from that use which, as we will see, connects the informativeness of ISOs to their representational primitives, domains of knowledge, and final recipients. Firstly, we maintain that ISO’s informativeness [II] interacts not only with the actual representational primitives, but also with their variability over time (§ 4). Secondly, we discuss the correspondence between representational primitives and domains of knowledge of those ontologies (§ 5). Finally, we explore the possibility of an epistemological discrepancy between human beings and software systems on the understanding of ISOs’ contents (§ 6).

The present reflection stems from what we consider as being a huge gap in the debate around ISOs: the inclusion of II among ISOs’ general aims has not been underpinned by a problematization of what II specifically consists of. However, since there is still no specific debate on II, the aim of this paper is neither to provide a final point on the matter nor to offer a new research paradigm to the field. The desideratum is a more modest one: arguing that discussing the implications of the current use of II (within ISO’s literature) on the assumptions underlying ISO’s definition may constitute a first step to fill the gap.

On Information System Ontologies

The Encyclopedia of Database Systems describes ISOs as follows:

[1] ISOsFootnote 1 define a set of representational primitives with which to model a domain (of knowledge). The primitives are typically classes, properties, and relations (among class members). The definitions of such primitives include information about their meaning and constraints on their logically consistent application (Gruber, 2009).

Of course, there is no shortage of alternative descriptions of ISOs,Footnote 2 whose role in the Semantic WebFootnote 3 architecture reflects, at least, two different purposes:

[2] to facilitate data sharing, inference, interoperability, aggregation, and integration on the Web;

[3] to support communication and mutual understanding between human beings, human beings and software systems, software systems (Goy & Magro, 2015).

Concerning [1], the definition has the merit of drawing attention to two focal points on which our investigation on ISOs' informativeness focuses on, namely

[4] the domain to model/systematize;

[5] the representational primitives for hierarchically and relationally modeling the domain.

Related with [4], another point concerns the aims of ISOs, with respect to which the Encyclopedia of Database Systems further specifies that

[6] ISO denotes an artifact that is designed for a purpose.

Such a purpose

[7] defines the domain that an ISO aims to represent,

[8] can vary from ISO to ISO.Footnote 4

[7] and [8] allow us to highlight one of the main differences with respect to the aims of philosophical ontology [PO], regarding which the plurality of hypotheses and methods of investigation (Runggaldier & Kanzian, 1998; D'Agostini, 2002; Varzi, 2011) does not prevent philosophical ontologists from maintaining that:

[9] PO aims to study the [9.1] whole of reality, by providing an (exhaustive) inventory of all there is (or might be)Footnote 5 (Berto & Plebani, 2015) – an aim that does not exclude the possibility of regional ontologies, the purpose of which is to establish [9.2] what there is within the domain of a specific discipline (Casetta, 2008).

This means, while PO's domain generally concerns the whole of reality or, at least, some of its specific sub-parts,

[10] ISOs' domains are arbitrary: that is, ISOs are in principle open to any domain of knowledge at any level of granularity, as well as being able to deal with anything that each ISO intends to represent.

On ISOs' informativeness

Within the debate around ISOs, the notion II has progressively assumed a fundamental role, to the point that some authors include it among the general aims of ISOs.Footnote 6

Just as for other (related) notions (such as completeness, accessibility, and repeated applicability), the investigation on II generally concerns two different (and related) horns of ISOs' analysis: ontology and meta-ontology. The former mainly refers to the entities populating an ISO as well as their systematization among ISOs' representational primitives.Footnote 7 The latter deals with the explicitation of methodology and theoretical choices for ISOs' design.Footnote 8

Providing a shared meaning of II can be, however, problematic, due to the absence of a specific debate on the topic. Without claiming to fill the absence of a shared meaning or to exhaust the potential of a debate that is still in the course of developing, the current useFootnote 9 of II seems to suggest that:

[11] an ISO is informative if and only if it is understandable for its final recipients.Footnote 10

But if so, on the basis of [1], II concerns

[12] the representational primitives for modeling the domain (see also [5]),

[13] the domain to systematize (see [4]) that is, in turn, related to the purpose for which an ISO has been designed (see [6]).

On Representational Primitives

Within PO's debate, the study of being is often associated with the individuation of primitive categories (Thomasson, 2019; Westerhoff, 2005) providing:

[14] the structure within which all there is is ultimately systematized (Norton, 1976; Rosenkrantz & Hoffmann, 1991),

[15] the identity criteria for classes of entities (Dummett, 1981; Frege, 1884; Lowe, 2006).

Such primitive categories – regarding which the various positions at stake primarily include objects, properties, kinds, attributes, relations, tropes, facts, events, processes, and so forth – can be related to each other both in terms of combination (Cumpa, 2019) or hierarchical organization (Chisholm, 1996; Grossmann, 1992; Hoffman & Rosenkrantz, 1994; Tegtmeier, 1992).

Compared to PO (see [14]), things do not change that much for ISOs, within which representational primitives supply the structure where ISOs' contents are ultimately categorized (see [1] and [5]). However, while PO's debate has generated different and incompatible positions on the individuation of primitive categories, classes, properties, and relations (among class members) represent the main and shared examples of representational primitives for ISOs.Footnote 11 Classes, which can contain sub-classes and/or be sub-classes of other classes, are sets of (class) members that share common features. Properties describe the various features of a class and of its members. Relations represent the way in which both classes and (class) members interact with each other.

Now, [12] specifies that II concerns the systematization of ISOs' contents, a systematization that [1] relates to three representational primitives: classes, properties, and relations. The fact that ISOs' debate identifies, at present, such primitives does not exclude that their list will never change – due, for example, to the advancements of IT/computer science in ISOs' design. This means, II is currently related to classes, properties, and relations, but II could also involve, in future, other representational primitives. In other words, [12] remarks the fixedness of the connection between II and representational primitives, but not between II and some specific primitives. Therefore, while representational primitives can change, what should not change, according to [12], is the relation between II and representational primitives – a relation that prescinds from the individuation of such primitives.

On ISOs' Domains (of knowledge)

[13] connects II to the domain that an ISO aims to model. [10] specifies that ISOs' domains are arbitrary and [8] adds that such domains can vary depending on the ISO in question. But [8] also explains that the purpose for which different ISOs have been created can vary from ISO to ISO.

[16] This means not only having different ISOs for different domains (of knowledge), but also different ISOs describing the same domain with a different purpose.

To account for [16], let us suppose to build an ISO that models the contents of a sweet box.Footnote 12 This means, the sweet box represents the domain of our ISO. Now, in modeling such a domain, we might be interested in the total number of sweets, classify such sweets in accordance with their different kinds (gummies, caramels, liquorices, and so on) or with their properties (color, shape, and so forth), or maybe our interest might concern all the items (number, kinds, properties) listed before. All these systematizations are, in principle, perfectly licit: their effectiveness will be measured by the purpose of our ISO. A parent who scolds her/his children after binging on sweets will be interested in the number of sweets left in the box, compared to the confectioner producing those sweets who wishes to know which kinds of sweets are the more successful. In this sense, both the parent and the confectioner will regard the same domain with a different purpose. Accordingly, changing our purpose might reduce or even nullify the effectiveness of a specific systematization, just as it increases the effectiveness of others, and vice versa.

Now, according to [1], ISOs model domains (of knowledge) by means of representational primitives. But what are such domains? Smith (2003) seems to suggest that

[17] outlining an ISO's domain means providing an inventory of (all) the representational primitives which populate such a domain.

If so,

[18] ISO's domain would coincide with ISO's representational primitives.

However, [16] specifies that different ISOs can describe the same domain with different purposes. And this is the reason why the confectioner of the previous example might not see, in the ISO modeling the contents of a sweet box, the primitives representing the more successful kinds of sweets. But then, in this case,

[19] ISO's domain would not coincide with ISO's representational primitives to the extent that there is something, of the domain, that such primitives do not represent.

All this considered, [18] and [19] constitute two contrasting positions on ISOs' domains, two positions that, on the basis of [13], might affect the notion of II. Indeed, by following [18], there is nothing but ISOs' representational primitives that the final recipients need to properly understand ISOs domains (of knowledge). Conversely, by following [19], ISOs' primitives could not suffice to provide the domain that an ISO aims to model. So, II would require more than the primitives.

On ISOs' Final Recipients

By maintaining that ISOs are informative if and only if they are understandable for their final recipients, [11] also relates the notion of II to ISOs' recipients. But, who are those recipients? According to Goy and Magro (2015),

[20] ISOs' final recipients can be both human beings and software systems.

And this is not surprising to the extent that [3] specifies that one roles of ISOs is to support the mutual understanding between human beings and software systems. [3] does not, however, imply that human beings and software systems necessarily understand ISOs' contents in the same way.

Tambassi (2021c), for example, maintains that the difference consists in the fact that, while human beings may have access (at least in some cases) to the references of ISOs' representational primitives in the real world, software systems, unless they are equipped with artificial perception, may not.

Goy and Magro (2015) also claim that

[21] ISOs' aims can change depending on whether such an ISO intends to support communication and mutual understanding between [21.1] human beings, [21. 2] human beings and software systems, or [21.3] software systems.

More precisely, ISOs mainly aim to share the definitions of representational primitives in the case of [21.1], to enhance the access of ISOs' contents (on the web) in the case of [21.2],Footnote 13 and to facilitate semantic search and interoperability as well as at enabling automatic inferences in the case of [21.3].

All this, however, does not exclude the possibility of an ISO that is informative for human being and not informative for software systems. In other words, the same ISO could be, at the same time, both informative and uninformative, depending on its final recipients. But then, on what basis is an ISO informative? When is the same ISO informative both for human beings and software systems? And then, should such an ISO be considered as informative for all human beings and all software systems? Considering the possibility of different kinds of II would allow ISOs to be informative for some specific final recipients and not for all of them. However, such a restriction should be, somehow, specified in [11]. Conversely, maintaining [11] without any restriction would entail that II is open to any kind of final recipients, complicating the process of ISOs' design.

Concluding Remarks

The absence of a shared meaning and of an actual debate on II makes it difficult to develop an exhaustive reflection on the topic. For this reason, this paper should be interpreted as an investigation aimed at capturing some theoretical issues emerging from the current use of II within the ISOs domain – a use which is summarized by [11] and which connects, indirectly, the notion of II to ISOs' representational primitives, domains, and final recipients.

Regarding representational primitives (see § 4), it has been shown that the fixedness of the relation between II and representational primitives does not correspond to the fixedness of the actual primitives (namely classes, properties, and relations), which could change according to, for example, the advancements of IT/computer science, without compromising the horns of such a relation.

As for ISOs' domains (see § 5), the issue is represented by whether such domains correspond to representational primitives. In case of correspondence, the representational primitives are all that the final recipients need to understand an ISO. In case of no correspondence, ISOs' domains are something external to the primitives, so that, understanding an ISO would mean, for the final recipients, to know both the primitives and the domain to model.

As for ISOs' final recipients (see § 6), the notion of II seems to require some further specifications to ensure that the same ISO cannot be, at the same time, both informative and uninformative, depending on whether its final recipients are human beings or software systems.

To conclude, we cannot fail to remark that the arguments presented so far do not even intend to exclude, at least, other two possibilities.

[22] The first one is that maintaining [3] – that is, ISOs generally support the mutual understanding between human beings, human beings and software systems, software systems – does not necessarily mean that each ISO must pursue this goal.Footnote 14

[23] The second one is that the notion of II specified by [11] does not imply that such a notion will never change nor that it will always represent a general goal of ISOs' design.