Knowledge and place-based development – towards networks of deep learning

ABSTRACT The influential work by Barca on place-based development, which has permeated policy and academic discourses alike in recent years, builds on the premise that localities are expected to utilize their endogenous potential rather than placing their trust in redistributive policies. This endogenous potential involves local knowledge and place-based knowledge, and how these two types can tap into actions. This has barely been explored in a systematic and comparative manner. This paper therefore examines 20 urban and rural development actions across Europe in order to understand how, and the extent to which, local knowledge and place-based knowledge are mobilized (or not). It makes use of empirically informed evidence to identify evolving mechanisms and to analyse how learning loops are triggered. We argue that it is crucial for leading actors in such development actions to pay attention to these different mechanisms of mobilizing these two types of knowledge and how to trigger learning loops. Since this analysis also highlights a number of shortcomings and inhibitors regarding the extent to which these collective knowledge and learning capacities actually inform actions over time, the concept of ‘networks of deep learning’ is suggested as a knowledge management principle for key actors in local governance.


Introduction
The aim of this paper is to analyse different ways of mobilizing local and place-based knowledge within the confines of local governance practices. It therefore aims to develop further our understanding of the role of learning networks, including interaction between local and external actors. Where actions, projects and programmes across different aspects of local development governance are concerned (e.g. spatial planning, environmental protection, economic growth management, etc.), there seems to be general consensus among scholars and local development actors alike on the value of including all relevant stakeholders, as well as the local populations who will ultimately be affected (e.g. Healey 1997;Innes and Booher 2003;Davoudi and Cowie 2016;Fazey et al. 2013;Lowe et al. 2019). However, a recurrent problem is how to mobilize and include these actors and their knowledge in ways which are legitimate and transparent, sustainable in the long term, and go beyond tokenism or merely giving them the opportunity to have a 'voice' in one-off dialogues, surveys or consultations. If local knowledge and place-based knowledge are taken seriously, the challenge is first how to mobilize these types of knowledge, and then how to use the outcomes of this mobilization to form a basis for continuous learning in a variety of development actions. In order to further develop the understanding of these issues, in this paper we suggest the concept of 'networks of deep learning' as a knowledge management principle. Moreover, the relevance of this challenge is underpinned by a growing discussion in policy circles regarding the importance of broadening the knowledge base for policy interventions, especially in terms of place-based development strategies (Barca 2009;Barca, McCann, and Rodríguez-Pose 2012; see also UN-Habitat 2009, 2017. However, research on how local knowledge and place-based knowledge are actually mobilized into different types of action is rather limited (Fazey et al. 2013). Although there is an understanding of the situatedness of the roles of local knowledge and place-based knowledge in each individual action (e.g. Lowe et al. 2019;Davoudi 2015), few studies juxtapose, compare and systematically analyse how these types of knowledge are mobilized in practice. In short, the 'toolbox' of local and place-based knowledge mobilization has not been systematically inspected or discussed. The present study therefore contributes with a comparative, 'trans-situational' analysis in order to reveal and to discuss how local knowledge and place-based knowledge in different local development actions and contexts are mobilized and integrated (or not) in networks of differently positioned actors. It draws on insights in the literature on policy mobilities (e.g. McCann and Ward 2011;Temenos and McCann 2013), which is especially relevant when discussing interactions between local and external actors, i.e. how policy ideas may travel between places. The following research questions acted as a guide: How are local knowledge and place-based knowledge mobilized and expressed? How are local knowledge and place-based knowledge incorporated into actions? How do these two types of knowledge flow between actors?
The analysis is empirically driven, and is based on 20 case studies of local development actions across 11 European countries. The case studies are all based on a common methodology (Weck et al. 2018;Weck and Kamuf 2020) grounded on fieldwork (involving interviews, focus group meetings, observations, workshops and textual analyses) in the respective localities, and were carried out between 2017 and 2019 within the framework of the Horizon 2020-funded project RELOCAL (Resituating the local in cohesion and territorial development). For the present paper, the case studies were analysed progressively by the authors, who took turns in a deep-reading process. In this way, the data was coded for original, rather than repeated information, noting the new information each case study brought to the analysis. In other words, there was no attempt to list similarities between cases, but rather to use an explorative approach to trace variations in how local knowledge and place-based knowledge were mobilized, and the forms of interaction between local and external actors.
The results point to a need for a better understanding of how local knowledge is generated, how it is related to place-based knowledge, how both, local knowledge and placebased knowledge, circulate between actors, and how these two types of knowledge form the basis for learning loops which develop over time. We argue that these learning loops go beyond knowledge exchange (Fazey et al. 2013), as they are based on long-term interactions. We propose that leading actors should strive for forming 'networks of deep learning', which can help them to sustain the mobilization, generation and integration of local knowledge and place-based knowledge, and as such, inform and shape the governance of local development actions. Here, 'networks of deep learning' refers to relations between various actors which are to be sustained over time, which positively condition mutually beneficial interaction on a common issue, and through which continuous learning takes place in terms of seeking and implementing sustainable solutions.
The next section provides a short review of the literature in the field, focusing on networked learning and drawing on insights from the policy mobilities literature. The aim of this section is not to explore the very broad social science literature on what knowledge is, the various roles and types of knowledge or the relation of knowledge to action or agency. It centres instead on the situations where knowledge is exchanged and learning can take place, with a focus on the linkages, or networked learning opportunities, between various actors. The two sections following this detail the empirical analyses, including making a distinction between local knowledge and place-based knowledge. The former relates to the local population's lived experiences, whereas the latter involves place-based professional experiences and expertise. The first section focuses on how local knowledge and place-based knowledge are mobilized, and the second on how, and to what extent, they are incorporated and trigger loops of learning in the analysed case studies. The paper ends by synthesizing the main observations and by further outlining and reflecting upon the characteristics of 'networks of deep learning' as a knowledge management principle.

On the role of knowledge in local development actions
As noted in the introduction, there is hardly a subject area in planning, development, or urban and regional studies where the role of local knowledge or place-based knowledge has not been highlighted as an important feature. Similarly, the involvement of various stakeholders and end-users, and the 'activation' of their knowledge, are generally regarded as aspects of 'good governance' (Davoudi and Cowie 2016;UN-Habitat 2009, 2017. Local governance and development projects typically relate to complex phenomena, and involve a number of different actors representing diverse interests and who are generally positioned differently (McFarlane 2006;Young 2016). As Kooiman (1993, 4) pointed out a long time ago, 'no single actor, public or private, has all the knowledge and information required to solve complex, dynamic, and diversified problems'.
Local development actions therefore often transgress disciplinary boundaries and require the active engagement of actors with different backgrounds, interests, professions, capacities and resources, and potentially also with different loyalties. Ultimately, however, all actors within an action, a project or a programme should work towards the same goals. The knowledge required for local development actions therefore centres as much on creating common understandings and fruitful working and learning relationships between actors (Innes 2004), as it does on knowledge about the details of governance processes and practices in different fields. Furthermore, reflexive and collective learning practices are important in order to make the best use of the local resources at hand, including the potential of local knowledge (Healey 1997). In short, the sensible handling of collective knowledge processes is a key feature of 'good' local governance (Innes and Booher 2003;Healey 2010).
However, Barca, McCann, and Rodríguez-Pose (2012, 147) argue that knowledge is often not readily available but 'must be produced anew [our emphasis] through a participatory and deliberative process involving all local and external actors'. Rather than thinking in terms of single actors, and of knowledge as something static which can be 'possessed', there is a need for a continuous process of knowledge creation in development actions through interaction between concerned actors. As Davoudi (2015, 323) notes, '[k]nowing is distributed and collective. In a unified account of knowing and doing, knowing is not a separate category; it permeates social relations. It is a socially constructed understanding that emerges from practical collaboration'. Similarly, Rydin (2007, 52-53) emphasizes that knowledge is 'generated in knowledge networks encompassing sets of relevant linkages' between social actors. The required knowledge is formed between the actors who have a stake in the action under consideration, ranging from the knowledge of the end-users to centrally positioned decision-makers. It includes individual learning, but more importantly also collective reflection within these networks. Knowledge, and generating new knowledge, is thus better understood as a collective, interactive and processual undertaking. However, how these 'sets of relevant linkages' (Rydin 2007) are anchored in concrete practices and interactions is an open question, as is the way in which they actually work to mobilize, integrate and engage knowledge for a common cause.
Related to knowledge is the concept of learning. Gilardi and Radaelli (2012; see also Wenger 1998) highlight the importance of reflexive social learning in relation to governance networks. They consider reflexive social learning to be related to dense socialization within networks, and the way norms and ideas are interpreted, shared and contested. Similarly, Schmitt and Van Well (2016) emphasize the role of individual learning processes in local or territorial governance processes, which are inevitably set in motion in joint projects and actions where inter-personnel networking and trust are central drivers, along with the degree of motivation and passion of individual actors. In the perspective taken in the present study, the idea of learning in networks is central to involving not only local people and their knowledge, and how this knowledge interacts with the expertise and experiences referred to here as place-based knowledge, but also the possible experiences and expertise in other places, or at different scales. It is argued that the role of networked extra-local interaction is crucial for local development (Cabiddu and Pettinao 2013), and this is often realized by involving non-local actors and their expertise. In a similar vein, Barca, McCann, and Rodríguez-Pose (2012; see also Barca 2009) strongly underscore the role of external actors in supplying ideas, models and programmes, and to some extent also management, arguing that an 'external elite' needs to replace local elites if these have failed. This obviously raises the question of whether external experts, often with agendas of their own, have sufficient understanding of locally situated issues and the local 'tradition' (Lennon 2017), and hence whether 'their' ideas would 'fit' into various local contexts. However, Robinson (2015) argues that it is primarily local policy-makers that 'arrive at' what policy ideas to welcome and to import, rather than mobile policy ideas sweeping in without any relation to the locality.
In any case, involving external expertise implies that ideas have to travel from one place to another, and this is, in fact, the focus of the recently emerged literature on policy mobilities. This field involves in-depth studies and theorization on how (policy) ideas travel between places. Learning from non-local actors through social networks should not be conflated into a straightforward copy and paste approach. Instead, ideas and beliefs on how to do things change or mutate as they are circulated between actors across space (McCann 2011;McCann and Ward 2011). In addition, their implementation is subject to 'barriers' of various kinds (McLean and Borén 2015; see also Stein et al. 2017;Schmitt 2020). Temenos and McCann (2012) argue that for new ideas to enter a place successfully, the place has to be prepared; it needs a certain institutional 'fix' in order to be able to receive, learn and implement new ideas and knowledge. Similarly, Dzudzek and Lindner (2015, 391) argue that policies do not just 'land' in a place, but interact with the local governance regime in 'a process of intense mutual engagement'. Moreover, Borén and Young (2020) emphasize the various informalities involved in networked learning. Although formal 'informational infrastructures' (McCann 2011; Andersson and Cook 2019) such as conferences, meetings and similar social events, are important in facilitating the transfer of ideas, knowledge and beliefs, recognizing that much learning takes place by serendipity within rather informal contexts is an additional key to understanding how learning in networks actually comes about.
These insights are pertinent in terms of considering learning and developing knowledge within networks across space, especially if external ideas or elites are given a strong role. von Schönfeld et al. (2019) argue that learning will take place the minute people start interactingit is a fundamental human quality. Fazey et al. (2013) point to the role of 'satisfaction' in engaging individual actors in the networks. Satisfaction is about positive affirmation of the people involved, and their emotional involvement. 'A reasonable level of satisfaction', as Fazey et al. (2013, 26) suggest, 'may mean that participants are willing to continue to share and exchange knowledge, which is vital for viable longer-term sustainability of a project. ' Lowe et al. (2019, 35) foreground that 'interactional' expertise is a key element in successful knowledge exchange between external experts and people who have local and practical knowledge. Interactional expertise, or knowing how to interact in order to promote knowledge flows between various actors, is important to facilitate mutual learning. Collaborative practices therefore appear to be central to triggering learning loops. However, von Schönfeld et al. (2019) also introduce a word of caution by pointing to the inherently optimistic narrative in the planning literature on learning by social interaction. They argue that local knowledge is framed within a positive story in which inclusion and collaboration are central and are seen as a remedy for a range of problems, particularly in terms of legitimizing interference in local places. One important aspect of collaborative local governance is that it is difficult to encourage people to engage, even if there is an explicit commitment to interacting with the local population in a certain development project. In addition, if local people become engaged, a related and wellknown problem is that it is seldom a representative sample of the local population.
Following Lennon (2017), we argue that it is important to understand the local perception of the public interest, since it seems to be crucial for the success of local development actions to relate to the local context and traditions, which also implies the inclusion of local knowledge and place-based knowledge at the respective localities. However, as von Schönfeld et al. (2019) also note, interactions may also lead to learning outcomes which oppose the set goals; hence simply interacting is not a guarantee of success. This claim also further underlines the need for critical inquiries into 'actually existing' (cf. Brenner and Theodore 2002) learning interaction and networks.

Mobilizing and incorporating local knowledge and place-based knowledge
This section discusses a number of significant mechanisms for mobilizing and incorporating local knowledge and place-based knowledge by drawing on a number of evidenceinformed features from 20 case studies from 11 European countries (see Table 1). In order to unpack the information in the respective case studies on networks, as well as the linkages and interactions involved, a distinction is made throughout the paper  local knowledge and place-based knowledge. We argue that this distinction is important for a better understanding of learning networks. It stems from a typology of knowledge within local development actions (Borén 2020), and makes it possible to differentiate between lived and professional experiences of a place. Local knowledge is conceptualized here as stemming from the lived experiences of a place, and the knowledge a local population acquires through working and living in a place. Place-based knowledge is the knowledge of a place that bases upon being professionally engaged within the local governance of that place. It therefore also relates to the professional and institutional experiences of (local) governance practices, referring primarily to the knowledge of local practitioners or policy-makers operating in municipal administrations, local development agencies or consultant companies. This distinction is important, since the two types of knowledge cover different aspects of knowing a place, and both are crucial for the successful implementation of development projects.

Identifying local knowledge and/or place-based knowledge
Since the practice of knowing is distributed and collective (Davoudi 2015, 323), the question arises as to how local knowledge and/or place-based knowledge can be identified within local development actions. Examples from the case studies analysed include a number of different types of investigation, such as dialogues, interviews and surveys, which had been further processed to become relevant knowledge for the actors leading the action in question. Examples include 'interactive community consultations' (UK 2) and a quantitative survey on local inhabitants' perceptions of district management and the implementation of projects for urban regeneration (RO 2). However, in both cases the results played only a minor role in the further process of local action. A different observation is reported from the case study in the Balaton Uplands (HU 3), where a questionnaire survey was undertaken to mobilize the local youth and to incorporate their interests in the local development strategy (Kovács and Nemes 2019, 21). This example shows how information was not only gathered, but also processed as transferable and applicable forms of knowledge to inform strategy building. The case of the 'Stockholm Commission' (SE 1) illustrates ways in which the 'search for' and 'incorporation of' local knowledge is facilitated by approaching groups which are generally regarded as very difficult to engage with, such as young people or migrant women. These people were actively sought in relevant neighbourhoods in the southwest of Stockholm. A number of young people were trained in interview techniques, and then employed to conduct interviews with other young people. A total of 186 interviews were conducted and analysed, and the results were discussed at workshops with developers who were planning to build new apartments in the neighbourhood. These activities also involved 'translation', in the words of Rydin (2007) and McFarlane (2006), to construct a set of general principles, or 'keys', to facilitate this knowledge being transferred and used by other actors in developments across the city (Borén 2019).

Inclusion and exclusion of local knowledge and/or place-based knowledge
The comparative analysis of the 20 case studies also indicates that an effective way of mobilizing knowledge within an action is to involve actors with local and/or placebased knowledge in key positions. This becomes most obvious in terms of what are known as EU co-financed LEADER actions (here FI 2, HU 3 and UK 1), which are based on the idea of involving specifically 'local' actors with insights into 'local' affairs and challenges. The three cases show clearly how the role of local knowledge is built into the design and delivery of strategies, decision-making and resource allocation of each action. The following quote from the case in the Northumberland Uplands (UK 1) illustrates how the Board of Members, consisting of local representatives, incorporates local knowledge by involving local people in key elements of the action: These meetings are held in the evenings, to accommodate the employed organisational representatives who are part of the Board of Members, and last from two to three hours. Here a considerable breadth and depth of local knowledge has been observed in operation, as it is used to support applicants to improve their explanation of a project, and/or increase the viability of the project design and likelihood of its acceptance for funding. (Brooks, Shucksmith, and Madanipour 2019a, 30) In the case of the 'Stockholm Commission' (SE 1), the action was driven by experts with extensive place-based knowledge, who made suggestions about what could work to meet the challenges in deprived neighbourhoods. Similarly, the action in the 'Homelessness Project' in Lewisham, a Borough of London (UK 32), was informed by place-based, rather than by local knowledge, as local-authority officers and ward-level council officers specifically influenced the action (Brooks, Madanipour, and Shucksmith 2019b, 25).
Overall, local knowledge and place-based knowledge are difficult to codify, since the social relations of a place and the 'sense of place' are complex individualized issues experienced, learned and developed through being in a certain local context over a period of time. These two types of knowledge, whether originating from laymen or planning professionals, are not easily made visible, nor are they easily transferable to other actors involved in the action in question (Rydin 2007;Davoudi 2015). A potentially useful way of resolving this problem might be to include individuals with either local knowledge or place-based knowledge, ideally representatives of both types.
In accordance with Temenos and McCann's (2012) observation that the locality has to be prepared in order to accommodate new policy ideas from the outside, cases where place-based knowledge was acquired externally (e.g. EL 1, NL 1, PL 2, see Table 1) indicate that a particular 'fix' has to be in place. This fix can be created, such as in the case of the Goth Village (PL 2, see below), but new actions can also be adapted to fit the current fix, adjusting the new to the old rather than the other way around. An illustrative case involves the Pas-de-Calais Mining Basin. Here, a non-profit organization, 'EURALENS', was installed to function as a 'local project incubator' and as a 'metropolisation laboratory'. The aim was to foster innovative local initiatives which would encourage the social and ecological transformation of this formerly highly industrialized steel and coal area (Blondel 2019).
A specific way of including individuals and their local knowledge involves the active participation of local citizens in initiating the action. In the case of the 'Participatory Budget for Lodz' (PL 1), this type of action can be understood as mobilizing and incorporating local knowledge into local affairs in a very direct sense. Specifically, the rural cases in the Northumberland Uplands (UK 1) and in Ostwestfalen-Lippe (DE 1) underscore the importance of incorporating local knowledge and place-based knowledge in actions, as the following quote illustrates: I think rural areas do really know well what their problems are. The chairman of the sports club knows it, the chairman of the heritage society knows it, the mayor of the municipality knows it, the councillors know it, the regional management knows it, and the district of Lippe knows it. We need less work on concept development. Consultant agencies are the only ones who benefit from that. (Matzke, Kamuf, and Weck 2019, 24) However, cases were also identified in which local knowledge was for the most part intentionally excluded. One example is the case of 'Induced Earthquakes' in the Northeast Groningen Region (NL 1), where the action was mainly driven by non-local external actors. Although local actors and their knowledge were mobilized as mentioned above, the resulting information was not really used in the action. As a result, the Groningen case is characterized by strong local resistance to the leading actors driving the action, and attempts to voice concerns in terms of the issues at stake (Trip and Romein 2019). Furthermore, in the case of 'Rotterdam South on Course' (NL 20), considerable local knowledge was excluded, though there were at least some opportunities for local knowledge to trickle in through direct interactions between local individuals and leading key actors. Unlike in Rotterdam, local individuals in Groningen spontaneously formed groups to mobilize local knowledge in order to make their case and influence the action. As such, these movements were effective, as it seems that they led to an improvement in incorporating the wishes of local people. Comparing these two cases from the Netherlands illustrates how commitment to the locality as a place differs remarkably between the two.
Several of the cases analysed from Hungary and Romania illustrate how Roma people were not included or were often underrepresented in terms of contributing their local knowledge (here RO 1, RO 2, HU 1, HU 2, see Table 1). The action in Encs District (HU 13) provides an illustrative example of this: The 'playing events' organised by Malta [the leading actor in the action] […] aimed to approach marginalised groups, such as Roma, through informal situations, and elicit their knowledge and voice through informal discussion.
[…] not including the findings of these public forums in the Micro-Regional Mirror is evidence of the exclusion of these marginalised Roma, for whom the programme had been initiated originally. Not a single Roma has ever been asked anywhere about what she/he wants, what she/he is in need of. […] In this sense, Give Kids a Chance [name of the action] failed to transform local institutions in a way that would empower local Roma with a voice to advocate for a more just distribution of services through participatory institutions. (Keller and Virág 2019, 26) In other cases, rather mixed forms of exclusion and inclusion could be identified. In the case of Volos City (EL 1), for instance, it was reported that the local knowledge of the scientific personnel working in the municipality was utilized effectively, together with a number of internal studies and reports. On the other hand, the knowledge of local research institutes was not sought, even though there was a department at the university in Volos which had conducted commissioned projects on similar issues for other municipalities and regions in Greece (Petrakos et al. 2019, 18). Clearly, the question of exclusion/inclusion of local knowledge and/or place-based knowledge is tightly linked to the use of power in deciding whose knowledge is considered relevant (Davoudi 2015). However, exclusion, as the case in Encs District (HU 13) illustrates, can also lead to a need for strategies to avoid reinforcing prevailing local lines of conflict between different social groups.

Shaping opportunities for interaction
In most of the cases analysed, special efforts were made to bring actors in different positions into contact with each other. In some cases, this constituted an important foundation of the action itself (e.g. DE 2, FI 2, FR 1, see Table 1). Overall, creating platforms for knowledge-sharing and discussing strategies, with the specific aim of including local actors and local knowledge, is a common way of mobilizing knowledge. Another illustrative example is the case in the Northeast Groningen Region (NL 1), where a platform was created to empower the locality in relation to external actors. Although this platform later became less effective, it triggered the formation of local action groups, which were often set up autonomously by concerned citizens. A few of them joined forces to form an umbrella organization, which also included housing associations, farmers and employers, thus bringing various types of expertise to the network (Trip and Romein 2019). Special platforms were also developed for the 'Monistrol 2020 Strategic Plan' (ES 1), so-called 'reflection groups' which brought together actors to contribute to developing the action, based on their understanding of the place, their situation and their prospects (Ulied et al. 2019).
Meetings of various kinds, whether one-off or systematically arranged over a longer period, were found to be a very common tool for mobilizing knowledge across almost all case studies. The knowledge generated at these meetings varied, ranging from information sharing to more open discussions which had the potential to trigger mutual learning between actors representing differing interests. A number of examples were also detected of how key actors mobilized local knowledge through direct contacts with local individuals, rather than in a (public) meeting. Such examples included face-toface meetings or other types of direct conversations with the general aim of incorporating the resulting information into the overall action. In this respect, an illustrative example is the case of NULAG Leader in the Northumberland Uplands (UK 1), where the current programme officer used to develop an independent relationship with applicants through phone conversations and home visits.
On two of the three occasions when the researcher attended a NULAG meeting, it was observed that the programme officer was able to use this knowledge to correct or moderate board members' expressed assumptions and partial knowledge. (Brooks, Shucksmith, and Madanipour 2019a, 29) Similarly, the action known as 'Youth Centre' in the German town of Görlitz (DE 2) illustrates how representatives of the leading actor approached adolescents by organizing weekly meetings in a central square in the city in order to invite them to participate in social activities (Kamuf, Matzke, and Weck 2019, 21). These ways of shaping opportunities for interaction and ultimate knowledge exchange are obviously time-consuming, and are therefore likely to be costly. On the other hand, first-hand contacts with individuals who have a good command of local knowledge are most likely to be beneficial in developing common understandings, and can give local people the feeling that their contributions are valuable. However, the case study on the 'Urban Regeneration' initiative in György-telep, a neighbourhood in Pécs (HU 14), shows that contacts which are too close can be problematic.
The dwellers got used to the permanent presence and the availability-even on weekendsof the social workers, through whom they could get help to solve their problems at almost any time. However, this type of relation also created a kind of dependency on the social workers, which we call "informal paternalism". (Jelinek and Virág 2019, 19) In the case of Rotterdam (NL 20), the leading actor, the local office of the National Programme Rotterdam South, decided not to engage too much with local interest groups in terms of acquiring local knowledge. Instead, they communicated with local individuals in a more arbitrary way, only when it was considered relevant to the action. This approach offers more flexibility than planned or even continuous interaction, but inevitably inhibits transparency in terms of which, and whose, knowledge is considered valuable.

Triggering learning loops
Thinking in terms of 'loops' indicates an interest in how knowledge is turned into practice, i.e. how knowledge circulates and how it is developed, adjusted, mutated and consolidated among actors (e.g. McFarlane 2006). Analysing learning loops involves tracing the knowledge flows that bring in new knowledge from other places (e.g. by including external experts) as well as the mechanisms that are at play to spread and store information (e.g. such as different types of media).

Learning from other places
In some of the cases analysed, it was clear that the action should draw on experiences from other places which had undertaken similar actions. For instance, the action 'Stockholm Commission' (SE 30) was explicitly informed by a similar approach in Malmö (SE) (Borén 2019), and the action 'EURALENS' (FR 17) built on a somewhat similar policy model to that developed by the international building exhibition Emscher Park in the German Ruhr region in the 1990s (Blondel 2019). Other actions specifically mention learning from other places as an important source of knowledge. For instance, in the action 'Induced Earthquakes' (NL 19), the establishment of a platform for dialogue between actors was inspired by the experiences of a similar platform created in relation to a major infrastructure investment at Schiphol Airport in Amsterdam (NL). Similarly, the case of the state planning agency 'EPA Alzette-Belval' in the Lorraine-Luxembourg Border Region (FR 2) facilitated knowledge exchanges with other EPAs [Etablissement Public d'Aménagement] in France as well as within the EGTC (European Grouping of Territorial Cooperation) Alzette-Belval (Evrard 2019).
Learning based on experiences from other places, or on similar actions in other places, is expected to encourage reflection on how to adapt initiatives to the local context, and not just make a blueprint copy of the organizational schemes or strategies involved in other actions (e.g. McCann 2011). The action known as 'Community-led local development' in Kotka Town (FI 2) provides an illustrative example of this. It was connected to a network of organizations, stakeholders, practitioners, experts and decision-making bodies not only within, but also beyond Kotka. These informed how the action could adapt to changing conditions throughout its implementation (Fritsch et al. 2019a, 25).
In terms of facilitating learning loops, learning from other places is evidently supportive, but it cannot replace local analyses of what to do and how to do it (Robinson 2015). For instance, in the case of 'Overcoming Fragmentation' in Volos City (EL 1), diffusing the 'Ecosystem of Collaboration' was a priority for local leaders. The intention was to fuel it by establishing a network of 13 Local Development Agencies across the country, at the expense of local engagement with the action (Petrakos et al. 2019, 15-16).
Another way of triggering learning from other places involves organizing field trips. In the case of the Northumberland Uplands (UK 1), funding was made available to send potential applicants for rural development projects on study visits to develop their ideas through contact with lighthouse projects in other parts of the country, and even to engage in knowledge exchanges with various rural projects in other European countries (Brooks, Shucksmith, and Madanipour 2019a, 30).

The role of conventional and digital media
The media also plays an important role in learning processes and in involving the public in place-based development. Media coverage makes actions known, and can reach people who might wish to engage in activities and contribute to them with their specific knowledge. How the media represent places also contributes to common understandings and local identities. Conventional and digital media alike can offer an arena for debating and gathering opinions, which can be crucial in developing a common understanding of the action under consideration (e.g. DE 2, FR 1, see Table 1). In addition, the media may be crucial when the action is dependent on applications from the population, as in the case of Lodz (PL 1). However, there it was noted that the various outreach formats (e.g. leaflets, website, posters and social media) and support activities (e.g. mobile and stationary advisory services, marathons for writing proposals) were not sufficient to engage people sufficiently. The Northumberland case (UK 1) also noted a lack of reliable and deep-reaching forms of communication, which were considered crucial for attracting new grant applications to keep the action alive. The traditional means of advertising in local newspapers was said to be in decline, among other things, due to fewer retail outlets for newspapers in this large, sparsely populated area. On the other hand, a shift to digital media was not yet applicable due to poor broadband connectivity (Brooks, Shucksmith, and Madanipour 2019a, 27).
Other cases, such as Lieksa (FI 1), involved an extensive digital media strategy for developing a strategic plan to make decision-making processes more transparent. One part of the digital media strategy was to make possible for residents to give feedback to the municipal administration (anonymously or otherwise) through an electronic form. In addition, meetings of the City Council were recorded and could be watched online. Another tool for facilitating learning loops involved improvements in the city administration's communication practices, including frequent bulletins and engaging more heavily in social media. As a side effect, it contributed to a deeper understanding among the residents of the various rationalities underpinning municipal decisions (Fritsch et al. 2019b, 21).

External expertise and the role of academic scholars
Often the involvement of 'external' experts (e.g. in the form of consultants or academics) is viewed critically in local development actions, since they do not necessarily have sufficient local knowledge or place-based knowledge when they begin the work they are contracted to undertake. This can hamper their ability to make use of their expertise in the local setting (Lowe et al. 2019; see also Lennon 2017). However, a different approach was undertaken in the action in Maslomecz Village (PL 2), where mutual trust and social relations were built up over years between the village association and archaeologists as external experts. As a consequence, the results of their archaeological excavations, which took place over a period of 25 years, could be turned into a key development asset for the village . Similarly, the case of EURALENS (FR 1) illustrates the argument made by Barca, McCann, and Rodríguez-Pose (2012) concerning the role of external elites. These are actors that may be external to an action (and not only to the locality), who introduce fresh ideas from which local actors can benefit if they are receptive to them. In Monistrol de Monserrat Town (ES 1), for instance, the whole action was coordinated by a 'Territorial coordination team' led by an external consultant who was not caught up in local relations, unlike the case of the Balatan Uplands (HU 3) where a local NGO was commissioned. This team in Monistrol de Monserrat Town 'was responsible for managing the participatory plan, gathering local knowledge, structuring a narrative embracing existing and new initiatives, and validating proposals' (Ulied et al. 2019, 22).
It is also worth noting that some projects made active use of generic academic knowledge which was not specifically place-based. In the action in Stockholm City (SE 30), for example, the key actors commissioned research on various topics in order to inform the action more broadly (Borén 2019). In the case study 'Induced Earthquakes' in the Northeast Groningen Region, academic expertise from different academic disciplines was utilized to better equip the locality in question with evidence-informed knowledge on a number of aspects. This helped them to present arguments against external actors pursuing a contrasting agenda (Trip and Romein 2019, 25). The learning loops which emerge from engaging with academic knowledge can also infuse the action with legitimacy from 'outside', in the sense that academic knowledge is considered impartial (Lowe et al. 2019). On the other hand, introducing academic knowledge that is not regarded as impartial may be problematic and lead to mistrust and frustration, as shown in the case of Groningen (NL 1).

Conclusionstowards networks of deep learning
This paper has presented an empirically informed analysis of how, and to what extent within local governance arrangements, local knowledge and place-based knowledge are mobilized, integrated and trigger learning loops. To this end, it has focused on the forms, expressions and ways of mobilizing local and place-based knowledge, and the learning loops involved when different actors interact to promote and implement a local development action. The study's empirically grounded distinction between local knowledge and place-based knowledge sheds light on two types of knowledge related to a locality: the lived experiences of a place and the professional experience of it. This distinction helped identify a number of key mechanisms used by different actors to mobilize knowledge or even trigger learning loops between actors and institutions. We argue that being aware of this distinction and the various mechanisms is important for understanding how the two types of knowledge can be captured, utilized and expanded within the action under consideration. Our analysis also highlighted the contemporaneity of these two types of knowledge within the actions in question, which is another important issue for further consideration in research and practice.
The study of the different mechanisms for mobilizing knowledge helped distil a number of different categories of inclusion and exclusion. The former is related to tacit ways of incorporating knowledge which cannot be codified or measured. The analysis of mechanisms which led to exclusion in the case studies demonstrated that the integration of knowledge is always selective, and that there are forces at work, intended or otherwise, which regulate what and whose local knowledge or place-based knowledge is deemed to be relevant. The analysis also produced evidence of cases in which the exclusion of knowledge can lead to frustration and mistrust, making processes opaque and even reinforcing existing lines of conflict.
The organization of learning loops is central to the flexibility and adaptability of local actions. The case studies indicated local and regional variations in terms of how social knowledge relations were organized across Europe, and the extent to which knowledge was promoted to and from different actions and their (leading) actors. It is important to reiterate here that leading actors in local development actions need to be mindful of these different characteristics in terms of how knowledge is mobilized, how it flows and how it actually leads to learning loops. However, the analysis also highlighted a number of shortcomings and inhibitors regarding the extent to which these collective knowledge and learning capacities actually inform actions over time.
For this reason, we want to suggest the concept of 'networks of deep learning' as a knowledge management principle for leading actors in these and other local development actions. We argue that leading actors could be expected to ensure the inclusion of both local knowledge and place-based knowledge in development actions, and that constructing mutually beneficial learning loops based on these is a collective and relational effort, demanding opportunities, time and ultimately money. In a sense, a network of deep learning is an 'informational infrastructure' (McCann 2011), but one which involves 'satisfaction' (Fazey et al. 2013), social learning (von Schönfeld et al. 2019) and opportunities for informal exchanges (Borén and Young 2020). Moreover, in a network of deep learning, the 'linkages' (Rydin 2007) are sustained over a long period to sufficiently include, and make possible the learning of the local traditions (Lennon 2017). The latter is especially important if external experts have a key role.
Moreover, leading actors need to acknowledge that such social and collectively formed networks of deep learning must be carefully established and maintained, but also that they require a degree of flexibility in terms of changing contexts or participants. Most importantly, however, leading actors should be aware of the above-mentioned contemporaneity of the two different types of knowledge that circulate within these networks. The current analysis understands a network of deep learning to consist of relations ultimately based on trust between people with different capacities and backgrounds (and loyalties), and these relations need to be mutually beneficial or meaningful for all actors involved.
Moreover, networks of deep learning need to be goal-oriented in a sense that goes beyond the mere procedural dimension of local governance and instead aligns with local 'traditions' and understandings of what locally constitutes the public interest (Lennon 2017). Our cases have indicated that there is a need to overcome tokenism and lip service to local populations, and therefore to mobilize local knowledge and place-based knowledge in networks of deep learning can provide alternative ways to organize the deliberations in order to advance the governance of local development actions. Thus, the crucial task is not only how to utilize and incorporate these collective capacities in an action, but also how to build trust and transform the inter-personnel and inter-institutional relations into long-term opportunities for continuous learning. Ultimately, sustaining the relations depends on their role as socially meaningful channels for the flow of knowledge between actors.
The concept of networks of deep learning further suggests that leading actors should identify ways of incorporating local knowledge and place-based knowledge into learning loops, as these actors are in a position to inform and further shape local development actions. It is this relationality between differently positioned 'internal' and 'external' actors, and the mobilization and cross-fertilization of local knowledge and place-based knowledge, which may trigger learning loops within such networks of deep learning. Hence, in research and practice, there is a case for further exploration of how networks of deep learning can develop, how they eventually wither away, and what can be done to form, stabilize and utilize them more effectively in local development actions.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
Research for this article was funded by the European Union's Horizon 2020 research and innovation programme, under grant agreement No. 727097, project RELOCAL (Resituating the local in cohesion and territorial development), 2016-2021.