1This chapter consists of feedback, which highlights the process of collective and interdisciplinary development of evolutionary models of settlement systems, but also and especially the diverse perceptions of those engaged in TransMonDyn. Whereas chapters 14 (‘Between models and narratives: transitions in settlement systems’) and 16 (‘Ontological points of view on the transitions of settlement systems’), coordinated by modellers for whom ‘modelling’ constitutes an object of research in itself, and who favoured theoretical insights and increasing generality, this chapter proposes to place the practical dimension at the heart of its discussion. The chief concern is to return to the points which have challenged the knowledge and practices of a group of researchers, who will be distinguished, for the sake of simplicity, under the names of ‘thematicians’ and ‘modellers’ in this enterprise of sharing knowledge and producing models together. What have been the main stumbling blocks or difficulties to overcome? What is the weight of individual experience (of scientific background) but also of the modes of collective work in the impressions of the varied participants? How was the work of acculturation to the approaches of the ‘other’ carried out? We are the better able to engage such questions because both the thematicians and the modellers were already familiar with the practices of the other community – something which has facilitated exchanges and contributed to the success of the project. This feedback thus ‘empirically’ sheds light on the process as a whole – a combination of numerous micro-processes –: the point here is to focus on the adjustments and readjustments linked to the numerous interactions which took place during the TransMonDyn experiment (figure 1).

2The points that we will focus on are, on the one hand, the difficulties linked to the identification of the transitions to be modelled, on the other hand, the question of the place of empirical data in the different phases of elaboration of the model; we will then return more specifically to the role of modellers as possible mediators. The discussion will be punctuated by reflections on other difficulties encountered, such as those stemming from habits of thought regarding privileged dimensions – the temporal dimension for the historians and the spatial dimension for the geographers – or indeed that of the level of specialisation of the researchers involved, whether in thematics or in terms of operating mode.

3In the framework of this presentation, we will be referring to the appendixes, which express individual points of view: those of Pierre Garmy, ‘thematician’, on his confrontation with the modelling procedure, and of Alain Franc, ‘modeller’, on questions of the choice of formalism, as well as that of Arnaud Banos, who proposes a synthesis involving the different ‘thematician-modeller’ interactions which are at the centre of the co-construction.

Identifying a transition: a subject for discussion

4To begin with, the notions of regime and transition, which require a sequential perspective, constituted an unaccustomed framework for historians, archaeologists, linguists and geographers (who played the role of ‘thematicians’) used to understanding the processes they study within a temporal continuum.  Historians, however, are accustomed to dividing time, in the framework of the periodisation that they establish in order to account for the transformations they identify. As for geographers, they pose these questions with regard to space, knowing that a place can never been considered as a ‘closed system’. In any case, these chronological and spatial divisions are objects constructed on the basis of an interpretation – a point, moreover, that leads Grataloup to parallel the artificiality of continents and of periods1. In the framework of TransMonDyn, the difficulty has lain in the capacity to trace the contours of regimes considered as stable: that is, the changes which have affected them are not of such a kind as to modify the spatial configurations of settlement systems, unlike the processes at work within the transition.

5The problem posed by the identification of stable states explains why, starting from the first list of transitions established collectively, some of the researchers engaged in studying a transition have adjusted the initial framework, whether to propose two transitions (in transition 3 – see chapter 6) or to modify the chronological framework, for instance by dissociating transitions which were originally supposed to succeed each other without a gap (transitions 5, 6 and 7). As an example, the narrowing of transition 6 (termed Romanisation) to the chronological sequence from the second century B.C.E. to the first century C.E. (instead of the second century C.E.) is due to the fact that the sharp decline in the number of establishments in the second century C.E. may be considered as indicating a reconfiguration of the settlement system, hence as a new transition, rather than as a simple change in the state of the system (cf. chapter 2). Transition 7 (‘From the world of antiquity to the medieval world, fourth to eighth centuries’) provides another example of the redefinition of chronological frameworks linked to the work of modelling: in order better to grasp the processes at work, the authors analysing this transition felt impelled to schematise a state R₁₊ and thus to define two states within regime 1, an initial state corresponding to a third century inherited from the Early Empire (R1) and a second state (R₁₊) corresponding to the initiation of the transition2 in the course of the fourth century.

6Each transition has been the object of a general description, formalised according to a generic conceptual framework developed during the first seminars (cf. chapter 2). The question of narrowing the focus arose in a second stage because each group favoured one dimension during the modelling of the transition, and not all the dimensions considered during its description. This step resulted from an interaction between ‘thematicians’ and ‘modellers’ in defining those salient facts in the processes observed that could lend themselves to modelling. For instance, transition 6 is centred on modelling the interactions between Roman authority and Gallic elites, while transitions 7 and 8 focused on the inter-relationship between the different centres of power and their spatial implications. As for transitions 1 and 2, they placed at the centre of their model the colonisation movements involving new spaces. In this framework, the interaction with the modeller was the determining factor, a point to which we will return below.

7The identification of the transitions has benefitted from the exchanges which took place during the seminars. These were essential, in that they made it possible to clarify the conjuncture between two successive transitions. Thus, following an exchange between the authors dealing with transition 7, the thematicians of transition 8 made it clear that the modelling did not concern the dismemberment of the Carolingian empire but rather its effects on a local scale (displacement of the competition between aristocrats and shrinking of their spatial span, a growing conflictual tendency and militarisation in society, territorial control at the local level and increased pressure on the peasants). Regarding the final version of transition 7, it plainly emphasised the emergence of new poles of power in the context of competition between Frankish and Visigothic kingdoms, and of the Church’s increasing power. In both transitions, it appeared more clearly that the dismemberment of the empires constituted the context, and not the object, of the modelling. Discussion with the researchers of transition 7 thus permitted a better definition of transition 8, the first steps of whose modelling had lacked a definition of the granularity of the processes.

8This aspect is fundamental. Generally, the formalisations are single experiments, which must be connected to a context and an objective, and which are therefore difficult to compare. Here the fact that it is not only the results but the processes of construction that are compared, for two transitions with closely related frames of reference, made it possible to render choices tangible and to establish links between the stages.

9These first exchanges benefitted from identification of a common framework of interpretation (cf. chapter 2), which allowed some transitions to be set side-by-side, even if they were not successive and not shared by the same collective – something which permitted the identification of similar processes in widely separated periods. Thus Jean-Luc Fiches, who worked on transition 5, has been especially interested in the processes at work in transition 10, which concerns the dynamics of settlement and urbanisation of South Africa.

10The sessions of collective reflection during the seminars also brought to light questions of point of view and contributed to exposing differences of appreciation concerning the objects and the processes modelled. Indeed, the scientific background and the specialities of the thematicians necessarily had an impact on the focal distance adopted in studying each transition. This has been particularly visible in the case of transitions which follow each other chronologically. Thus, the difficulties of relating transitions 7 and 8, which nevertheless deal with apparently similar processes (fragmentation of power and space, emergence of new poles) originate not only in the diversity of geographical contexts and spatial scales considered, but also in the specialisations of the thematicians involved (Antiquity for transition 7, the Middle Ages for transition 8). In the processes modelled, the researchers of transition 7 were concerned to locate the modifications of those elements considered representative of the functioning of the ancient system (fragmentation of the provincial and civic framework, emergence of the dioceses, territorial competitions, evolution of the urban network). They based themselves on the expertise of a specialist in the Early Roman Empire and focused on the changes in relation to the initial situation (R1). On the other hand, the thematicians of transition 8 focused on the central processes of medieval political and social organisation, the conditions of whose apparition they sought to grasp (polarisation of settlement around seats of power – churches, castles – involving the emergence of new central places and the formation of a new territorial mesh at a local scale – parochial territories, castellanies). They therefore privileged the situation at the end of the transition (R2) over the conditions of arrival of the new regime, whereas the researchers of transition 7 concentrated more on the conditions of disappearance of the initial ancient regime. With respect to transitions 5, 6 and 7, the relations between them were probably facilitated by the fact that they deal with the same space and that the thematicians all adopted – given their specialisation – a point of view centred on Antiquity.

11As for transitions 1 and 2, they were at first closely allied because of their thematic proximity, since both are concerned with the process of expansion of a settlement. The question of the colonisation of new territories (transition 1) was the object of an initial modelling, on the basis of which the question of the ‘expansion of a settlement in a space already inhabited’ (transition 2) was formalised. Thus, the work on transition 2 was undoubtedly influenced – indeed, constrained – by the framework conceived for transition 1.

12This initial step was principally due to discussions among thematicians, but the modellers seized the opportunity thus offered of ‘entering the narrative’ and sketching out grids for the comprehension and definition of objects. The meeting between these two cultures was nevertheless made more arduous by the fact that the ‘thematicians’, who have different cultures, could disagree with each other regarding their different ways of categorising (a discussion took place, for example, concerning the content to assign to the concept of ‘city’, depending on the chronological and geographical contexts). For their part, the computer scientists preferred to put in the same basket elements that function in a similar manner, without waiting for a definitive position on the frontiers between categories. These sessions of collective work thus were occasions to raise ontological questions concerning objects that cross periods (cf. chapter 16).

13In the same register of categorisation, the modellers tended very (too) rapidly to isolate a process by ‘unfixing’ it both spatially and temporally, a tendency contrary to the procedure of the thematicians, whose concern was always to refer the processes to an observable empirical situation. Sometimes it is specified that what is observed in a given place is not necessarily valid at the same moment for another space (thus, while they deal with phenomena that occurred within the same span of time, between 1000 BCE and 1200 CE, transition 2 – 1000 BCE/1000 CE - deals with the forest forager/Bantu relation in equatorial Africa, whereas transitions 5 – 600 BCE/400 BCE- , 6 – 200 BCE/100 CE - and 7 – 100 CE/600 CE – focus on southern Gaul, and transition 8 – 800 CE/1200 CE - refers to north-west Europe). The areas studied were in fact defined according to the data on which various participants have worked. This double ‘unfixing’ (spatial and temporal), influenced by the modellers to move towards more genericity, often received a poor welcome from the thematicians. It did, however, contribute to the process of acculturation of both. The modellers, notably, were thus able to better evaluate the importance of such and such a process or event, localised in time and space, as a result of the more or less vigorous resistance opposed by the thematicians to their attempts at simplification.

14This involves a crucial element, because the question arises of the generic nature of the processes observed, and therefore of the possible transfers from one transition to another. These sessions devoted to collective exchanges made it possible to identify all such reasons for friction (among thematicians, among modellers, and especially between thematicians and modellers).

15The exchanges that took place on the occasion of these seminars were essential for clarifying the objectives, the concepts and the procedures proper to each discipline. Thus, the field excursion to Villetelle (to the excavated site of Ambrussum, then to a prospected site) allowed the modellers, but also the non-archaeologists, to become aware of the artefacts with which archaeological discourse is constructed. In turn, the presentations of the different practices of modelling fostered the acculturation of the thematicians, especially those who had little knowledge of the diversity of conceptual frameworks of the modellers. On this occasion, certain thematicians also discovered discussions, even oppositions, among modellers, of which they sometimes found it hard to grasp the issues involved (the role of operating development environments, the relation to data … see below).

16The ontological procedure (‘regional ontology’, to take up the terminology of chapter 16), which allows for defining the objects, relations and properties of the transitions, could not be constructed at the outset. It has required a period of apprenticeship, the more so because it has been implemented at two levels – that of each of the transitions and that of the transitions as a whole in a comparative perspective (cf. chapter 16). This time of maturation could undoubtedly have been reduced, given a classical production framework, which would have placed the thematicians in the position of ‘demanders’ and the modellers in the position of ‘producers’. One of the interests of the TransMonDyn programme lies precisely in this procedure of the co-production of models, more demanding from the point of view of cross-disciplinary investment, but also more rewarding.

Data – at the heart of the relation between thematicians and modellers

17Besides the question of defining the object of the transition, what has aroused lively interrogations and perhaps mutual misunderstanding (see appendix 1) is the place of data3 in the modelling process (cf. figure 1). These intervene at different stages, although with differences depending on the transitions.

18In most cases, the data have served to identify the transition: for instance, in transition 7, it is because the epochs of the appearance and/or disappearance of the villae are known that one can, for the area under study, evaluate their number at each date and account for a sequence of states which provides information about the dispersion of the settlement grid between the second and fifth centuries. These data are the product of a sizeable effort of collection, and they attest that there was indeed a transition in the settlement system; they are at the heart of the narrative that the thematician offers to the modeller.

19In other cases, identification of the transition depends more on the interpretive models (the way in which the change is conceptualised), rather than directly on the data, even if these are obviously the basis of such models. In transition 2, which deals with the spreading of the Bantu populations, the narrative is grounded on paleo-climatic considerations for evaluating the environmental constraints and on linguistic and genetic observations for identifying the routes taken by the migrant populations. The data make it possible to validate the model, to a certain extent. On the other hand, the parameters of the model are abstractions, in that no data exist to determine their value. In the case of transition 8, centred on the processes of concentration and hierarchisation of settlement between 800 and 1200 in north-western Europe, the identification of the agents of the simulation model and the definition of their characteristics have been effected on the basis of the productions of contemporary historiography. The parameterisation of the model was carried out on the basis of the data available for the regional sub-group of Touraine. The exploration by simulation served to test the weight of certain of the parameters selected, and to define the value intervals for which the results of the simulation are in keeping with the historiographical model.  

20Once the transition is identified, the data intervene in the characterisation of the two states considered as stable systems (the two regimes) and of the phase of change (the transition). Their selection has been made in accordance with the common conceptual framework defined at the beginning of the experiment (cf. the five dimensions defined in chapter 2).

21The data are, in particular, a basis for setting out the processes in detail. On the basis of the corpus of data, the thematicians and the modellers discussed what one ‘knows’ or ‘does not know’, so as to determine how the gaps in information may be approached. Thus, for transitions 5, 6 and 7, if it is possible to enumerate and describe the places of settlement, one cannot set bounds to their areas of influence.

22The role of data is not limited to the preliminary phase. They intervene as well in the construction of the model and its implementation by computer. A portion of them are drawn on to describe the initial situation and elaborate the parameters associated with the processes being modelled. It is even possible that the application of the model may be the object of searches for new data, within the framework of the adjustment of parameters relating to new questions: thus, in transition 6, defining the payoff values for the Imperatores and the Gauls made it necessary to bring other sources of information in order to grasp the forms of logic at work in the interactions between the Roman generals and the Gallic elites.

23Data are also used in the phase of calibration and evaluation of the model. In the domain of the social sciences, it is rare to be able to assign an empirical meaning to the parameters and measure them with precision. Researchers will select the parameter values which permit reproduction of plausible results. In this phase, only the thematicians are capable of evaluating the relevance of the results in relation to empirical knowledge (see transition 6). Thus, the parameters associated with the metaphors of the ‘link’ (chapter 5) or the ‘resistance’ (chapter 12) cannot be directly measured, even if these notions make sense on the thematic level.

24The positions of the thematicians and modellers on this question of the use of data (appendix 1) bring to light their dissymmetrical knowledge of this dimension in the process of co-construction:

• The thematician is in the situation of someone with poor eyesight, in that he does not see the process of formalisation as a whole, whereas he is a major actor in the fundamental step that consists in passing from the ‘empirical domain’ to the ‘conceptual framework of the thematician’ (cf. chapter 16). He possesses a thorough familiarity with the object of the modelling, and sees the data that he has collected and supplied as the major input of the model. The schematisation necessary in the formalisation of the transition implies, however, a selection of data as a function of the processes identified, and hence a renunciation of translating the complexity both of the data themselves (heterogeneity, incompleteness, polysemy) and of the phenomena observed. As described by Pierre Garmy (appendix 1), this situation may engender a certain frustration on the part of thematicians, although modulated by their degree of familiarity with modelling. All admit that simulation does not aim at producing a detailed description of the object under study but rather a better comprehension of the processes at work in the phenomena observed.

• As for the modeller, he poorly perceives the process to be modelled, but he has experience in the overall process of modelling. His relation to the data is more distant, in that he has not collected them himself (moreover, he is not necessarily conscious of the ‘cost’ involved in acquiring these data). What counts, however, is the narrative constructed on the basis of the data, hence the result and not the building-blocks of the structure. He is nevertheless conscious of the fact that these elements will be put to use in several steps of the modelling process. What is more, when the data do not exist, he is led to find a solution by confronting the needs of the model with what the thematicians are able to provide: thus, in the framework of transition 2, it is thanks to an exchange with the modellers that Jean-Marie Hombert has proposed orders of size to quantify the groups of forest foragers present in the equatorial forest and the groups of Bantu likely to overcome the obstacle that the forest milieu constitutes for them (by circumventing it or by following corridors). The parameterisation of the multi-agent model simulating transition 8 (800-1200: concentration and hierarchisation of the habitat in north-western Europe) has also led the thematicians involved to supply many more orders of size that they would have wished. This phase of the operation, which confers a different status on the ‘data’ (observation and evaluation) could not have taken place if there had not been a reciprocal understanding of the issues associated with the data, on the one hand, and with the model, on the other.

The modeller – an objective mediator?

25Since the point is to report on the different experiences, it should be noted that the TransMonDyn project has permitted the creation of 10 collaborations, 10 co-constructions, which may be observed from the thematic point of view, but also from the point of view of their methodological application (figure 1). The model is a co-construction, and as such it constitutes a unique experience, as much for the learning that takes place during the process as for the product, that is, the model itself. If it is complicated for the thematician to free himself of his methodological habits (habitus), what about the modeller? Can he approach this experience without a methodological a priori?

26If one considers this co-construction as a pathway (an elaboration), the trajectory, reflecting negotiation between the actors, which leads to a point of convergence belongs to a sort of ‘possible domain’ bounded by the respective acquisitions of the thematician and the modeller, and to which the process of co-construction is attached: the domain of empirical reference and that of development by computer.

27It is in these two domains that the respective roles of each are most striking: thus, they constitute the input and the output of the co-construction. These methodological habits may intervene at both levels of the co-construction: on the one hand, in the definition of the conceptual framework relating to the thematic domain of reference, on the other hand, in the definition of the conceptual framework of modelling. It can nevertheless be difficult to dissociate these two stages, which are not so very distinct in the co-construction: this will be illustrated starting from the notion of the point of view adopted (the dimensions targeted by the model) and the selection of the formalism.

28With regard to the dimensions targeted by the model, transitions 6, 7 and 8, as previously seen, provide a good illustration of the diversity of choices and their link with the experiences of the persons involved. The transitions may be organised according to a continuum, centred on the modelling of the same dimension: competition between powers (transition 6); competition and co-existence between powers, with integration of the hierarchical levels of the exercise of power and their spatial inscription (transition 7); exercise of power by different actors, with their associated spatial logics (transition 8). One cannot attribute all these choices to the nature of each transition. Would the diversity of the previous experiences of the modellers not explain the differences in the choice of certain characteristics of the model? The integration of the spatial dimension, the choice of the level of modelling (aggregated or not, generic or not) depend very clearly on their domain of experience. The geographer-modellers are tempted to integrate the spatial dimension directly, while the non-geographers do not necessarily do this, something that permits them, from a certain point of view, to consider the process from a more generic perspective. Thus, in transition 6, the model developed by the modeller (not a geographer) derives from game theory, which does not include the spatial dimension in the choice of strategies, while, with an entry that might be comparable, centred on political powers, the formalisation proposed in the framework of transition 7 is of the multi-agent type. The framework of multi-agent formalisation encourages spelling-out of the spatial dimension and conduces, from this point of view, to a reduction in the level of abstraction. Similarly, depending on past experiences, the dialogue between modeller and thematician will be oriented towards either aggregated levels of modelling (the level of habitats, for example, or of social groups) or individual (the level of peasant households or lords). For instance, the targeting, in transition 7, of the process of co-existence of powers certainly arises from the compatibility between a marked aptitude and interest of the thematician to concern himself with political measures and the repeated experiences of the modeller in designing models at aggregate levels (system of cities). This observation could be applied to transition 8, where the domain of competence of the modeller, who is interested in residential choices in contemporary cities, is perhaps not remote from the level of modelling selected (that of individuals – lord, farmer) and from the mechanisms of spatial organisation at work (resulting from the movement of individuals).

29These choices definitely have an impact, too, on the way in which the processes of transformation are perceived from a thematic point of view: in the case of transition 7, the processes are perceived at an aggregate level; in the case of transition 8, they are presented according to a ‘bottom up’ logic.

30At issue here are observations that do not at all detract from the experiment, when one considers the model as an ‘intermediary’, as is the case in the majority of the transitions of TransMonDyn. To push the experimentation further would require the ability to experiment on the process of co-construction, and above all to evaluate and compare it in an objective manner.

31Concerning the choice of formalisms, the modeller is undoubtedly the one more engaged, and in a very pronounced fashion (as appendices 2 and 3 illustrate): the diversity of possible choices renders comparison difficult. It is nevertheless possible to differentiate those approaches in which the developmental environment pre-existed the co-construction, whether in a form already implemented (transition 1, transition 2) or in the form of a meta-model (transition 5), from those in which the formalism was chosen in the course of the process of modelisation (transition 8, transition 7 – if the adventure is pursued as far as the elaboration of a Multi-Agent System). In the case of transition 6, the choice of the formalism has even reoriented the initial question. Such a choice inevitably has an impact on the final result, even if it is a complicated business to evaluate the constraining force exercised by an explicit framework of pre-existing development, in contrast with the non-explicit framework of the habitual practice of the modellers. Indeed, the choice of the formalism, or formalisms, owes much to the practice of the ‘modellers’ (and to their ‘philosophy’ when it comes to formalism): hence transition 6 was approached by way of game theory (Alain Franc4), transition 7 by means of a graphic formalism and an explication of the formalism of the ‘agent’ type (Hélène Mathian), transition 8 through a Multi-Agent System (Robin Cura) after significant modelling work performed with Cécile Tannier. Finally, the type of model selected is also linked with the type of result expected and the means of arriving at it: thus the Multi-Agent System can simulate all the transitions but are not economical when it comes to stipulations if one chooses to develop a descriptive model. Game theory, for its part, requires defining in advance all the possible combinations of interactions instead of determining only the principles of combination5.

Figure 1: Schematisation of the process of co-construction of models developed in the framework of TransMonDyn.

Conclusion

32This feedback highlights the richness of a project which has placed at its centre the diversity of scientific backgrounds of the participants on both the thematic and methodological levels. Beyond the difficulties this has created, what stands out is the interest of the reciprocal acculturation thereby engaged between thematicians and modellers, which, by breaking out of the classic ‘client/supplier’ relation, makes it possible, in a procedure of real co-construction, better to grasp the benefits and the limits of the process of modelling. To pursue the experiment, a number of developments of the project might be considered. Hence, it would be interesting, for transitions straddling two canonical periods, to compare the modelling proposed by two groups of thematicians, each specialising in one period. Likewise, it would be advisable to test different formalisms on a single transition in order to evaluate to what extent the choice of modelling orients the results.

33In conclusion, we offer a reminder that this kind of innovative procedure has a place in the longer-term perspective on the interactions of computer science and the social and human sciences (see the initial rejection of the data-base management systems –DBMS – in archaeology, then of the geographical information system – GIS). With regard to modelling we could endorse a statement made by Jean-Claude Gardin6 at the conclusion of a colloquium on data-bases in the 1980s :

‘L'introduction des nouvelles technologies de l'information dans nos domaines conduit à donner le plus grand poids à des questions fondamentales qui n'ont rien à voir avec l'informatique, et dont peut-être on s'étonnera un jour qu'il ait fallu attendre l'ordinateur pour que nous nous les posions’.⁷

34The appendixes proposed below illustrate the diversity of positions. These texts reflect personal points of view acknowledged as such by their authors. They give an idea of the richness of the exchanges that took place during the seminars.