Demystifying facilitation: a new approach to investigating the role of facilitation in group decision support processes. In Group Decision and Negotiation. Theory, empirical evidence

. We believe there are still large gaps in our knowledge of the facilitator role in group decision support processes as exemplified by the use of group model building approaches, and these must be understood and “ de-mystified ” if use of these methods is to become more widespread. We use the design and analysis of delivering online group model building, based on foundations established by [1], to form a better understanding of the role of the facilitator. Our experimental configuration makes use of Group Explorer (Decision Explorer), a Group Support System (GSS) that is based on the Journey Making methodology [2-4], but delivered across the Internet in a distributed manner. The facilitator is thus no more or less visible and connected to the workshop as any other participant by way of its online nature. Data from a workshop delivered in this way is analysed and the findings discussed in relation to the following work; i) Callon [5] on translation and specifically how facilitation addresses the questions of problematisation and interessement [6], ii) the work of Hiltz et al [7] on distributed GSS and the problem of animating methodology, and iii) the “ death of the expert ” [8], which gets to the heart of investigating the role of the facilitator as an expert in methodology . We conclude by reviewing the question of one possible end-point of this work – the demise of the expert facilitator and the rise of a participant-led group decision support process model.


Introduction
Debates about the processes of Group Decision and Negotiation (GDN) generally focus on methodology, expertise and facilitation, often independently, but sometimes conflated. But on the rare occasion where they are held just so far apart as to bring forth insights on the need to explore the ideas further, the comments and conclusions appear all too apparent. We therefore seem no further forward in our understanding of the practices and processes that should be adopted in pursuit of improved GDN performance than we were following Eden and Radford's seminal collection of studies on group decision support for strategic action [9]. Noting the failure of interventions in the realm of management practices, Eden and others encouraged academics and practitioners to be wary of dismissing such interventions on a matter of principle, portraying failure as one purely due to implementation that necessitated more contextualized and nuanced consideration of GDN practices from a hard setting to a soft [10][11][12][13]. We suggest, therefore, that there remains a research gap in terms of the need for theoretically informed empirical work to reflect the complexities of different processes for GDN; in other words, to employ more holistic approaches to process performance that reflect the many different demands that may be placed on a GDN intervention; and to review the complex relationships that may exist between GDN context and performance.
In particular, in this paper, we still see a large gap in our understanding and knowledge of the facilitator's role and that this must be understood and "de-mystified" as we consider the transition of GDN practices to a soft setting, as exemplified by the use of group model building in Problem Structuring Methods (PSMs). Our research focus is specifically on participant-to-facilitator interactions. Our context and opportunity arises from the need to facilitate stakeholder groups through a process of problem structuring where these groups are increasingly geographically dispersed. We base this on the evidence of requirements for four projects where the authors are either advising on the use of PSMs, or are directly involved as facilitators, where the staging of workshops with participants attending in person is proving difficult.
Building on the work of [1] we follow the idea of "distributed interaction within a PSM process", but still see the workshop as an important component of the process, at least virtually. With the general improvement in the quality of network connections and collaboration tools, coupled with low-cost easy to access cloud-based compute infrastructure we believe the means for exploring this way of working is now technically feasible and methodologically justifiable, hence the reality of distributed Group Support Systems (GSS) as a means of implementing group model building in a PSM. Naturally, the distributed nature of stakeholder interaction e.g. in the case of a charity with stakeholders spread between the UK and Asia, is itself part of the problematic situation and we are sensitive of the fact that distributed interaction within the PSM process cannot be separated from this. The empirical work we report in this paper is an exploration of the issue of facilitation as we establish a working environment in which to conduct such online, virtual workshops. The data we analyse is collected from one of these online workshops where we have demonstrated the capability to problem owners in organisations we are working with and where the presenting issue is in fact the question of how to make this distributed engagement work.
We adopt an experimental setup that makes use of Group Explorer (Decision Explorer), a GSS that is based on the Journey Making methodology [2][3][4], but delivered across the Internet in a distributed manner. Consequently, the facilitator is no more or less visible and involved in the workshop as any other participant by way of its distributed nature. Our analytical technique is based on ethnomethodology [14], which has recently been used to good effect to understand the micro-process of decision-making in workshops (e.g. [15][16][17][18]). In so doing, we make the following contribution to the literature. First, we build on the foundations established by [1], to form a better understanding of the role of the facilitator in this type of setting. In particular, our attention is focussed specifically on participant-to-facilitator interactions. We theoretically position our work in relation to the following: i) the work of Callon [5] on translation and specifically how facilitation addresses the questions of problematisation and interessement [6], ii) the work of Hiltz et al [7] on distributed GSS and the problem of animating methodology, and iii) the "death of the expert" [8]. This perspective enables us to take a broader and nuanced view of expertise, which gets to the heart of investigating the role of the facilitator as an expert in methodology. Finally, data from one of the workshops organised to demonstrate technical and methodological feasibility in this distributed manner is analysed and the findings discussed in relation to our theoretical expectations. In particular we examine the question of the possible demise of the expert facilitator and the rise of a technology enabled and participant-led group decision support process model. The remainder of our paper unfolds as follows. First, we review the literature on facilitation in GDN, delineating the dimensions of facilitation, explain our theoretical underpinning, and then bring the two ideas together in developing our theoretical model. Second, we present the data and method we employ. Third, we present and discuss our results. In our final section we highlight our contribution to extant literature and suggest implications.

Literature review on facilitation
Classic work on facilitation follows the seminal work by [19]. Here the concern was on the facilitator as the 'helpful intervener'. Here, the intervener strives to improve group dynamics and decision making or provide a learning environment to help participants gain confidence of an interpersonal nature in order to help them transform the patterns of communication. Indeed, much of the work on GDN focuses on the facilitator that strives to influence situations toward desired goals in human activity systems in which they intervene. Here the facilitator attempts to balance what is to be done with how it will be achieved; see for example [1,12,[20][21][22][23][24][25][26].
The role we are defining for the facilitator is somewhat different. We begin to problematize the role as follows using group model building in a PSM engagement as the focus of our group decision support process. The question of facilitation seems to be situated within existing PSM practice, so the facilitator as a role is already there, has always been there in the process, and always originating from a methodological root. We imagine the written accounts of methodology as a product of first-hand experience in facilitating the methodology. There are many published methodologists, but they are all also practitioners. Is it therefore possible to theorise about PSMs without extensive practice-based experience? Problematisation seems to require us to break the bond between the roles of methodologist and facilitator and place our focus on deconstructing the latter; the thoughts of the methodologist are largely what we know already from the literature.
We can problematize the role in three ways; 1. Through isolating the facilitator by making them as on-par with a participant as possible, creating a laboratory to study facilitator interactions (online method) 2. By analysing case studies divorced from codified PSM methodology (and therefore the associated methodologist) i.e. deconstructing a non-codified case to tease-out the facilitation role (if any) (non codified case method) 3. By distilling the essence of the facilitator role from the bulk of the PSM and GDN literature. To a certain limited extent this has been done in our literature review here but there is perhaps some further concentration that could be performed to tease-out nuances. However, it is unlikely to produce anything exceptional by way of results (literature method) To some, the role of the facilitator seems tangled with questions of leadership and, worse, the notion of systems thinking [27,28]. Not as anything precisely defined, and associated with any particular methodology, but as a quality of an individual that uniquely sets them apart to take on the role of facilitator when dealing with the sort of messy problems that PSM engagements are designed. We suggest that this is an unprofitable line of enquiry as it is unlikely to lead to any widely useful contribution. Our focus therefore remains with the performative i.e.
what are facilitators actually doing when they facilitate a PSM engagement, and therefore preserve our theoretical underpinnings in ANT/Mangle [16]. Whilst it would be interesting to explore the analysis of the facilitator role in non-codified PSM use, precisely because it would be an analysis of facilitation by a nonmethodologist in a PSM-like setting, we defer this to future work.

Theory
Recent focus drawing on pioneering work of Keys on the sociology of scientific knowledge [29][30][31][32],and recently work by OR scholars drawing on socio-materiality, particularly the works of Latour [6,33,34] and Pickering's Mangle [34,35] attention has been paid on to important agendas regarding theory, behaviour and outcomes pertaining to (particularly Soft) OR processes, including facilitation. We note that these studies have recognized that interventions are both temporally enacted affairs and concerned with becoming coordinated practices through the performance of using models as objects, but the studies are not adequate in addressing in full issues relating to facilitation in interventions. Therefore, some significant methodological and epistemological challenges remain [6,[36][37][38]. Relevant to our work on facilitation is an extenuation of socio-materiality from Callon [39]. He outlines a number of themes which we feel are relevant to our study, in particular, the Co-Production of Knowledge Model (CKM). In the CKM he recognises a persistently enriched contest between the production of standardised knowledge and knowledge that takes into account the complexity of local circumstances or context [39]. In the space between the two is the problem of facilitation. Callon's CKM notes that the typical mapping of the problem structuring onto an expert-lay divide -in which experts possess expertise and participants possess local knowledge that can challenge assumptions made by those applying expertise to particular contexts. The experts as facilitator do not capture the capacity of participants to be involved in all elements of knowledge production. Nor does it challenge understandings of a problem that may be as highly differentiated as those in held by the expert. In the same way that experts question their understandings through practice, so must the participants. Thus, under CKM, knowledge is co-produced through a process of dynamic, collective learning involving those for whom an issue is of particular concern, whether as a result of their expertise, their personal position with respect to the problem at hand or their personal experience of the problem. This explicitly recognises more socially distributed, autonomous and diverse forms of collective problem structuring [1,40]. Problem structuring is no longer a property of expertise [8], and the knowledge it produces is no longer accorded special privilege over other knowledge. This process does not eliminate the need for the involvement of expertise, rather it removes its privilege and emphasises that it is, on its own, insufficient.
Callon's CKM is useful in capturing a theme central to debates over expertise in decision-making. Expertise is more widely distributed than many might imagine. The question becomes how to mobilise and to diversify that expertise and what happens to the expertise of those the facilitator during this mobilisation. By addressing this question new kinds of understanding may be generated that unsettle the taken-for-granted aspect of problem solving. Here, we introduce an experiment that explores distribution of expertise/facilitation to other people, things and places. To understand why this distribution of expertise is different.

Constructing the Experimental Setup
A standard Group Explorer installation was repurposed for deployment in a cloud-computing environment.
Group Explorer provides a 'wrapper' to the Decision Explorer software -which is "designed to record, analyze and present qualitative data argumentation relating to strategic policy issues and modelled as cognitive maps" [3] -such that multiple participants can share in the development of a cognitive map in a facilitated workshop.
The wrapper provides a web-based interface to participants and also manages the various phases of model development and the storage of data about its dynamic development to support the sort of analysis presented later in this paper. A sketch of the installation and deployment process for the Microsoft Azure cloud computing service is described in Appendix A. The motivation for choosing this type of internet-based hardware platform was to address the following needs: • To facilitate remote connection to the Group Explorer environment from any geographic location without having to deal with organisational firewalls and access limitations • To avoid procuring and maintaining hardware, thus shifting costs from capital to operational, consistent with many organisations' strategies to migrate their IT infrastructure to cloud services • To instantiate the software environment only when required for a workshop, thus further pushing operational expenses as low as possible by making best use of the pay-for-use model of the service provider • To enable migration of configured Virtual Machines (VMs) to higher processor and memory specification hardware should there be a need for increasing performance. The management and configuration tools from the service provider are specifically designed to monitor for performance issues and enable migration.
A single online collaboration tool was used to provide both the audio conferencing capability and the means to share the screen of the computer hosting the Decision Explorer modelling software. Feasibility was tested with both TeamViewer and Citrix GoTo Meeting products 1 . For the feasibility workshops reported in this paper the usual two facilitation roles of a Group Explorer workshop was replaced by a single facilitator, who was both facilitating the group and also controlling the modelling using Decision Explorer.

Conducting Experiments
Establishing the workshop would follow normal means of working up to the point where an online workshop is to be used. Having agreed to the use of an online workshop the technical means to join the workshop need to be managed carefully. To help in the process of demonstrating feasibility to clients three documents were prepared and circulated to participants a few days before the workshop was due to take place. In addition to a data collection permission form, based on a standard template we use for normal workshops, we also provided a

Data Sources
Our approach is based on recording and analysing participant-facilitator interaction during an online PSM workshop. The data consist of audio transcripts and the Excel Spreadsheet created from the SQL Server Export Wizard report generated from the workshop data held in the Group Explorer database. The two datasets are tied together by timestamps.

Data Analysis
Three online group model building workshops to demonstrate capability were held as follows:  Figure 1. 3 The initial issue explored was how to support asynchronous modelling too. The Group Explorer setup was left running all weekend, hence the date range, to enable issues to be added and connected in the model without the facilitator being present at the client console. Whilst it worked technically, it was decided fairly quickly that this mode of working would not be explored further. However, we already have requirements for distributed workshops that will involve stakeholders separated by many time zones and the methodological issues raised by periods of un-facilitated participation could thus be investigated with the same experimental setup. The second and third workshops were deemed successful enough to warrant the conclusion that the online capability was operational, that having demonstrated feasibility the approach could now be used for future client workshops.
The audio recording from the third workshop was analysed to demonstrate the feasibility of the approach for more detailed future analyses of facilitator-participant interaction. The start time corresponds to the announced meeting connect time, 15 minutes before the workshop was due to start. As can be seen below, 9m:16s of the 15 minutes allowed were required to establish attendance and connectivity. N event T start Description 1 0:00 Facilitator is greeting participants as they appear on the audio conference and dealing with questions. One participant asks if there is enough time to "go and make a cup of tea" before the start, which they then proceed to do.

Discussion and Conclusions
Our work has focussed attention on shedding more light onto a subject that has remained equivocal. The process of developing an online group model building capability for projects with widely distributed stakeholders has given us an experimental framework to investigate the problem of facilitation at a micro-level. The attention to practical development of capability that could entail the decentring of the facilitator avoids the trap or decent into the purely critical and keeps the work at an empirical level.
The viewpoint piece [43] suggests that the development of group decision support has been by a number of 'gurus' and reflects on their legacy and succession. As pointed out in the literature review this status of 'guru' is associated with the combined role of facilitator and methodologist, although it is mainly knowledge about the latter that is reported; the healthiness of the field is evidenced by the continual development of methodology without much or any reference to the role of facilitation. Recent work by [44] provides hard data that can be used to refute any suggestion of stagnation "When combined with other recent survey evidence, the use of PSMs and Business Analytics is apparently extending the scope of OR practice", but the question remains whether these hints of a problem emerged because of something lacking in the area of facilitation, or more specifically in the facilitator as the sine qua non of methodological knowledge.
Our review of theory suggests that the role of facilitator, as purveyor of methodology in decision making engagements, is just another form of expertise that can be critiqued and potentially decentred from the essential business of group decision support. Our preliminary experiments have been light on methodology, both in terms of explanation to clients or in anything particularly creative in methodological design. The use of Group Explorer with a simple modification to the conventional use of Decision Explorer, coupled with its delivery online via a cloud service and with the workshop glued together by a reasonably sophisticated audio conferencing and screen sharing system provided a lot of the scaffolding for the group model building. In effect by implementing a distributed GSS that could be considered pre-packaged and largely separate from the process of facilitation. However, from the point of view of Callon on translation [5], and specifically how facilitation addresses the questions of problematisation and interessement [6], it was still the facilitator that initiated the workshops and who was essential on the conference call to explain how the process would work.
In the extract presented in Table 1  unpicking how a methodology plays-out in a group workshop and further analysis is required to fully understand this. Whilst we appreciate this would help us to improve group decision support processes generally, and is a worthwhile and perhaps necessary task, we also admit to the following agenda inspired by ideas of the "death of the expert" [8]. What if through further research we could understand the role of facilitator sufficiently well so that it could be coded into a software platform like Group Explorer? Rather than being puzzled by the question of whether a PSM engagement functioned because of the skill of the facilitator or because of a property of the methodology we would have sufficiently separated the two to gain clarity that the question of function could be investigated solely as a property of methodology. Although of course begging the question as to meaning of function. For the purposes of current experiments and future work our meaning is simply that of whether the group decision support process started at all and led to decisions being made.
We acknowledge the limitations in our work. Our analysis centres on the methodological, procedural and expert role of the facilitator, especially as initiator of process and enabler of lay expertise, mainly from the broad perspective of ANT. This has been at the expense of detailed micro-analysis using theories of behaviour such as Activity Theory [15]; however this is further work that can be carried out now that the experimental framework has been made operational and the method of data collection simplified. An additional strand of work we envisage is to return to the question of facilitator as initiator and how this role comes about, and an examination of the trust that must come into being in the relationship between the client and the facilitator.
To conclude, in our new experimental setting the facilitator has been literally decentred, the visual clues of being the centre of attention in the workshop have been removed and the facilitator is just another voice on the conference call. What if the audio cues could be replaced by software cues, perhaps supported by rule engine?
This is speculation and perhaps where the development strands of GSS and PSM come together in a general group decision support process, but further understanding and de-mystification of the facilitator role may open the door to a proliferation of PSM/GSS application platforms. Whilst this may be technically feasible, this speculation brings us back to the essential puzzle of a PSM engagement; the initial problematisation and interessement [6]. Is this at all possible without a facilitator regardless of the properties of methodology and