CO-CREATING CHANGE:

CO-CREATING CHANGE:

EXAMINING LEARNING AS A CHANGE INTERVENTION

Salvador Dalí, Old Age Adolescence Infancy - The Three Ages (1940), Oil on canvas, 50 x 65 cm, Dalí museum St. Petersburg, Florida (USA).

Caption to the image ‘The three ages’ of Salvador Dali (front page):

The year 2009 is not only the year of my graduation, but also the year of the 395th anniversary of the University of Groningen. The university celebrates this with a range of activities referring to the theme ‘Arts Meet Science’. Art has always been an integral part of humanity's great quest for knowledge. The interchange of knowledge between artists and scientists has led to many of our most important advances. For a great part the surrealistic artists, as Salvador Dali, contributed to that. Their form of arts was inspired by Freud and Jung.

Sigmund Freud and his disciples Carl Jung discovered the ‘psyche’ or ‘soul’. Freud unwittingly rekindled an interest in the metaphysical realm, which science had shunned in its quest for knowledge. He then endeavoured to study it in the same way the physical level had been: By applying reason.

While Freud laid the scientific groundwork, Jung leaped forward in his exploration of how the unconscious reveals itself through symbols. In this respect, artists were needed to join the quest for knowledge. Jung himself painted and sculpted his dreams and visions so that he could better understand them.

"The creative process, so far as we are able to follow it at all, consists in the unconscious activation of an archetypal image and elaborating and shaping the image into the finished work. By giving it shape, the artist translates it into the language of the present and so makes it possible for us to find

our way back to the deepest springs of life." - Carl Jung

Inspired by the work of Jung, Salvador Dali wanted his work to be a link between the abstract spiritual realities and the real forms of the material world. Through his painting, he could bring the inner realities of the subconscious to the conscious mind. Often, his works provoked reactions of surprising, abhorrence or excitement among the viewers. He was convinced by invoking such violent reactions and affective experiences he would let people start to release ‘old believes’. Only then values and norms in society could change.

The thoughts of the surrealistic artists link with an amount of present change theories of organisational practisers and researchers. These authors and their works (e.g. the book ‘Presence’ of Peter Senge e.a., and the ‘U-theory’ of Otto Scharmer) were a great inspiration for me during conceptualizing the model in this thesis.

CO-CREATING CHANGE:

EXAMINING LEARNING AS A CHANGE INTERVENTION

Master thesis, Msc BA, specialisation Change Management

University of Groningen, Faculty of Economics & Management of Organisation 25th of May 2009 ANIEK J.B. GEERTS Student number: 1334336 Helmondseweg 98 5751 GG Deurne Tel: +31 (0)6 47426507 e-mail: Aniek.geerts@gmail.com Supervisors/ university: Dr. B.J.W. Pennink Dr. B.J.M. Emans Supervisor/ field of study

Drs. M. Muntinga Munthoff & Partners, Zwolle

CO-CREATING CHANGE:

EXAMINING LEARNING AS A CHANGE INTERVENTION

ANIEK GEERTS University of Groningen

We examined the role of deeper levels of ‘learning’ in the relationships between collective behavioural change and its predictors ‘new knowledge’, ‘interaction’ and ‘facilitation’ among two groups of policy directors at a Dutch municipality. One group was involved in a learning programme; the other group was not and functioned as comparison group. A significant difference was found in the scores on ‘co-creation of change’ (collective behavioural change) between the two groups. Contrary to our hypotheses, we found no mediating role of ‘learning’. However, interaction effects were found in the sense that when ‘triple-loop learning’ is high ‘new knowledge’ and ‘interaction’ both exhibit greater and positive relationships with ‘prototyping’, the second stage of ‘co-creating change’. We argued that ‘triple-loop learning’ could have a remarkable influence on ‘co-creating change’. The three existential stances on the human condition - knowing, action and feeling - are considered to be complementary components in the change process. This study presents one of the first empirical findings with a quantitative dimension in the wide volume of existing theory on (organisational) learning in relation to (organisational) change.

A theme commonly found in the literature on management is the need for concepts of organisational change that are suitable for an increasingly turbulent and complex business environment (Pawlowsky, 2003). In order to cope with the challenges of our time, organisations and institutions in all the key areas of our society need to develop the capacity to succeed.

Wierdsma, 2003; Metselaar, 1997; p. 29). Behaviour is an expression of competences, cognitive ability – intelligence – and personality traits (van der Maesen de Sombreff and Schakel, 1999). In order to create collective behavioural change, organisations commonly implement a range of interventions. Interventions implement changes in an organisational work setting, which improves organisational outcomes (Robertson et al., 1993). Looking at personality traits and intelligence as unchangeable (or at least hardly changeable) and as given states, these interventions should extend the competences of the employees to create change effectiveness by changing behaviour collectively.

However, research from Scott-Morgan (in Bastings et al., 2009) and Beer and Nohria (2000) demonstrates that 40 to 70 percent of organisational change projects to develop the capacity to succeed as an organisation failed last decade. According to Scharmer (2007), the reason for this is that people within organisations do not rethink the problems deeply enough, and people have to learn on other levels than they are used to. Through people’s awareness of each other’s thinking and the underlying patterns of that thinking, acting is exposed and can be changed in a sustainable manner by the people themselves.

This suggests that learning and changing are closely linked conceptually: the terms (behavioural) change and learning have very similar meanings. This view on organisational change has been expanded enormously in recent thinking on organisational learning. In terms of de Caluwe and Vermaak (1999) this is called green-print thinking. Although a volume of publications on the subject exists (e.g. Senge, Cummings and Worley, Wierdsma, Scharmer), there is lack of thorough empirical research in the area (Burnes, 2004). Tsang (1997) pointed out that one of the main reasons for this is that many of these publications on organisational learning have been written by practitioners and consultants seeking to prescribe and sell, rather than describe or analyse. Pawlowsky (2003, p.81) concludes that, there is a great variety of assumptions about the determining factors of organisational learning, but most of these are based more on a rudimentary theoretical framework than on empirical evidence, and therefore: ‘The dependent variables of the models on organizational learning clearly need more attention if the business and scientific communities are to understand the accepted economic necessity and value of learning and knowing’.

components must be considered complementary if organisational learning is to be promoted in order to change behaviour. Taking these three conditions and other literature (Scharmer, 2007; Wierdsma, 2003; Senge, 1999) into account, we determined three predictors of creating collective behavioural change – ‘new knowledge’ (knowledge), ‘interaction’ (action) and ‘facilitation’’ (feeling) – and examined the role of a deeper level of learning in the relations between the predictors and collective behavioural change. Based on that, we developed a model presented in Figure 1.

1 Conceptual model

We used a sample of 20 employees who where involved in a change intervention, and a comparison group who were not (N=20) at a Dutch municipality. This allowed us to empirically study the creation of collective behavioural change among employees by learning as a change intervention, which will give change agents better insights into how to manage the change process as a learning process within interventions to co-create change.

THEORETICAL FRAMEWORK Creating Collective Behaviour: Co-creating Change

New collective behaviour is developed from an organisational change perspective that organisations are a context within which people behave (Scharmer, 2007; Senge et al., 2007; Wierdsma, 2003). Consequently, organisations primarily change and act through their employees (De Vries, 2002). The proposed organisational change can be defined as the desired adopted behaviour of the employees.

The desired collective behaviour is dependent upon the content of the specific organisational change project. Scharmer (2007; 76-79) links the desired behaviour to the changing economic context. The shift of focus in how managers must approach their work mirrors a larger shift in the economy. The (Western) world has moved from a product-driven economy to a service-driven economy, and is now moving to an experience-, knowledge-, and innovation-driven economy. Each of these three economic contexts functions according to a different set of principles, and each requires a different managerial mind-set and relationship competence. Recent research and theorists (e.g. Scharmer, 2007; Wierdsma, 2003) state that, when the focus of value creation has shifted, the organisation’s relationship with the customer has evolved from push, to pull, to co-creating.

Although successful individual behaviour is not a guarantee for being successful together, successful individual behaviour is a necessary condition to perform together successfully (Senge, 1992). When individual behaviour must extend to the competence of being effective together, there is a need for collective competence. Competences are a vital condition of behaviour; competences are expressed in behaviour (van der Maesen de Sombreff and Schakel, 1999). Multiple definitions of competences exist, all of which contain shared elements. A commonly accepted definition is that, competences are the ability to transform insights, knowledge and skills in a specific organisational context into adequate and responsible acting (Kessels, 1999). With collective competence the organisation is able to co-create processes internally and externally. The external actors are the ones who review the organisation and judge its products or services; essential to co-creating processes with these external actors is that internal co-created processes are needed: co-creating change.

crystallised image. And the last stage, performing, we defined as achieving the institutionalisation of the crystallised image through practises and infrastructures (Scharmer, 2007).

New Knowledge and Co-creation of Change

Knowledge functions as a reference to make sense of how we act (Weick, 1990). Someone’s reality is not formed from a single perception, but from a construction of different images, stories and assumptions. People interpret experiences and knowledge by means by selecting, omitting and organising. Elaborating on the work of Weick, Pennink (2004, p. 130) developed a model in which constructed images of managers about the company, the environment, and the purpose of working together are related to the way managers act and steer. He states that these images are formed by processing and applying information to a true situation. Agrysis and Schon (1978: 10) state that: ‘All deliberate action has a cognitive basis, that it reflects norms, strategies, and assumptions or models of the world which claims to general validity…’ Therefore, new knowledge can function as an expedient to act or behave in a new way (Wierdsma, 2003).

To be effective together, Checkland (1981) argues that actors do not have to agree about how each actor sees the organisational reality. Although he argues that thinking and acting are intertwined, different views on the level of thinking and seeing the reality can be complementary on the level of acting and behaviour. The acceptation that people’s mental models are intertwined with their acting results in a view that acting is a way to get insights into reality. Therefore, in order to connect people’s realities to a focus on acting that will improve the organisation’s performance, knowledge sharing is required (Wierdsma, 2003).

H 1.1. Acquiring ‘new knowledge’ regarding the context of the change will lead to high

‘crystallising’.

H 1.2. Acquiring ‘new knowledge’ regarding the context of the change will lead to high

‘prototyping’.

H 1.3. Acquiring ‘new knowledge’ regarding the context of the change will lead to high

‘performing’.

The Role of Learning in the Relation Between New Knowledge and Co-creation of Change

Acquiring new knowledge is not a sole factor of influence on co-creation of change, however. Acquiring new knowledge is believed to influence learning. Learning is essential to the notion that people act on reflected norms, strategies, and assumptions or models. Members of organisations are not merely storage bins of past rational experiences, but interpreters of reality according to the specificities of their individual knowledge system (Pawlowsky, 2003). Employees socially construct the organisation as they continually gain new knowledge, and they use that to act and interact with each other and learn from those actions how to organise themselves for productive achievement (Wierdsma, 2003). This is a learning process in which employees are able to develop, test, and modify mental models or maps of organisational reality. Thus, the influence of acquiring new knowledge on co-creation of change is believed to be mediated by learning. This has been hypothesised as follows:

H 1.*: The influence of acquiring ‘new knowledge’ regarding the context of the change on

co-creating change is mediated by ‘learning’.

Interaction and Co-creation of Change

to the interaction between you and the others. If an interaction shift from one pattern to another can be detected, it usually involves all the participants in the conversation, not just a few of them.

Based on Scharmer (2007), in this study interaction is divided into conversation and listening.

The first type of conversation: Talking nicely. Conversations are the living embodiment of social fields, and they are an important starting point for improving social interaction (Scharmer, 2007). Isaacs (1999) describes four types, and the first type can be described as talking nicely. Operating effectively in such conversations requires the participants to conform to the dominant pattern of exchanging polite phrases with one another; not saying what is really on their minds (Scharmer, 2007). From a(n) (organisational) learning point of view, this type of conversation tends to result in dysfunctional behaviour; it prevents a group from talking about what is really going on.

The second type of conversation: Debating. The word ‘debate’ literally means “to fight or beat down”, which is exactly the pattern of this type of conversation (Isaacs, 1999). People use their arguments to beat or best their opponent, defined as anyone with a different opinion; talking tough to challenge the dominant view. The debate style can be useful in an intervention because it allows a group to get all the different views on a subject on the table. But, if the issue at hand requires a group to reflect on and change their habits of thought and guiding assumptions, as has been hypothesised in this study, yet a different type of conversation is needed: dialogue.

totally new basis from which to think and act (Isaacs, 1999). This is called generative dialogue; it is a time when genuinely new possibilities come into being (Scharmer, 2007).

The first type of listening: hearing what you already know. The first type of listening is hearing what you already know. With this type of listening, what people hear confirms their assumptions and mental models. The attention of listening is focused inward, on the people listening themselves.

The second type of listening: being surprised at what you hear. With the second type the attention is shifted from the person to his or her surrounding. During listening, people can deviate from their expectations and mental models. What they hear surprises them. It is listening through paying attention to facts and to novel or disconfirming data.

The third type of listening: listening emphatically. At the third type the listener tries to listen from the perspective of the other person and starts to see the world through the eyes of that other person or actor. As long as people operate from the first two levels of listening, the listening originates from within boundaries of people’s own mental-cognitive organisation. But when people listen emphatically, the perception shifts from the person listening to the other person, or to another living system.

The fourth type of listening: attention to the collective. During the fourth type of listening the person shifts the attention of listening to a deep level. With this type of attention during listening a collective space rises, and a collective experience develops in which all parties create a vision together and put into language what it is that they want to create. Where Scharmer speaks about a ‘collective space’ in which people have to let go of the patterns of their past thinking, assumptions and images to co-create a collective vision, Pennink (2004, p.137) speaks about an ‘image space’ where there are multiple possibilities to co-create a collective image. In addition to the possibility of co-creating a new image, Pennink adds two other possibilities: combining all existing images to create a collective image which offers lots of possibilities to vary, choosing one image to create a collective image.

Listening closely and communicating openly without any restraints - dialogue, being surprised at what you hear, listening emphatically and paying attention to the collective - ensure an awareness of (unknown) underlying patterns, which enable new behaviour. This has been hypothesised as follows:

The Role of Learning in the Relation Between Interaction and Co-creation of Change As is the case with acquiring new knowledge, interaction is believed to facilitate learning. Interaction is the gateway to each other’s knowledge (Scharmer, 2007; Keursten and Frijters in Rondeel en Wagenaar, 2002). Interaction is therefore also the gateway to each other’s and people’s individual underlying thoughts, assumptions, drives and values, and allows people to learn about their own and each other’s acting. When these underlying patterns become clear, people are able to get a clear view about the causes of people’s actions and behaviour. By having a clear view of the reasons why people act in a certain way, people can reflect on that and apply it to their own future acting. Therefore, the influence of listening and conversation on co-creating is believed to be mediated by learning. This has been hypothesised as follows:

H 2.*: The influence of ‘interaction’ on co-creating change is mediated by ‘learning’. Facilitation and Co-creation of Change

On paper a change process generally seems easy to deal with, however, a significant change can be expected to be a very difficult process for employees. Changing behaviour can be a difficult and painful journey for employees (Burnes, 2004). In change management literature, practitioners agree that facilitation is vital for successful behavioural renewal when continuously ‘tempting’ employees to change (Tjepkema et al., 2002). Therefore, facilitation is defined as: assisting the progress of the change process by making it attractive and easier for employees to cope with the difficulties they face during their changes in behaviour (Tjepkema et al., 2002).

determinants of motivation to exert themselves to change. Situational constraints include lack of proper tools and equipment, materials and supplies, budgetary support, and time (Noe et al., 2003). This has been hypothesised as follows:

H 3.1: A high level of ‘facilitation’ will lead to high ‘crystallising’. H 3.2: A high level of ‘facilitation’ will lead to high ‘prototyping’. H 3.3: A high level of ‘facilitation’ will lead to high ‘performing’.

The Role of Learning in the Relation between Facilitation and Co-creation of Change Just like acquiring new knowledge and interaction, facilitation is believed to influence learning. Employees of organisations need tools and practice fields to learn how to use new methods more effectively in order to accomplish their corporate and interaction goals. Educational and industrial psychologists have identified several conditions under which employees learn best (Noe et al., 2003). When employees know why they have to learn, have the opportunity to practise, and are able to keep up with the learning process it will positively influence the learning process. Thus, facilitation is believed to have a positive influence on learning. This has been hypothesised as follows:

H 3*: The influence of ‘facilitation’ on co-creating change is mediated by ‘learning’. A Deep Level of Learning and Co-creation of Change

place (Pawlowsky, 2003). Groups are where the individual’s view of the world is shared, mediated, and influenced.

From this perspective on learning, Agrysis and Schön (1978) suggest a terminology of three levels of learning: single-loop learning, double-loop learning, and triple-loop learning. Single-loop learning means that people reflect on their actions and the focus is on how to improve the status quo. According to Scharmer (2007), this level influences the change level of reacting. Double-loop learning goes one step further and includes a reflection on the deep, taken-for-granted assumptions and is aimed at changing the status quo. This type of learning implies an adjustment to the environment, and is based on the assumption that organisations have theories-in-use, interpretation systems, and frames of reference that guide and determine behaviour in organisations. If the organisation’s environmental feedback challenges the organisation’s assumptions, and if these assumptions are differentiated, redefined, or altered completely in order to fit the environmental demand, then one can say that double-loop learning has occurred (Pawlowsky, 2003). Triple-loop learning, also called deutero-learning, involves learning how to learn. This kind of learning requires collective reflection on how to govern organisational and individual values and drives. Garratt (1990) calls this learning type an integrated learning cycle, which requires a necessary state that he describes as a helicopter view. Deutero-learning is a construction of higher-order rules based on experiences and insight (Pawlowsky, 2003). According to Bateson (in Pawlowsky, 2003), Learning type III, analogously to deutero-learning (Pawlowsky, 2003), involves attainment of knowledge about the development and the meaning of habits, and is only rarely possible to do as an individual without interaction with others because it occurs only in religious and spiritual experiences or in psychotherapy. This assumes an emotional impact on the learner when reaching this level of learning.

A given learning type may be more appropriate under some circumstances than under others. Learning may well necessitate merely simple adjustments in order for norms and standard operating procedures to be defined, but for changes at the level of new organisational behaviour, Scharmer (2007) states double-loop or even triple-loop learning is acquired. This has been hypothesised as follows:

Figure 2 Conceptual model with operationalised variables

RESEARCH SETTING Research setting

The object under study is a group of policy directors (N= 100) at a municipality in the north-east of the Netherlands. Municipality Emmen created the vision to optimise the policy development process to meet the needs for the contemporary involvement of all actors. To achieve the objectives of this vision, the policy directors require an appropriate set of (new) skills and (new) knowledge. Therefore, the municipality set up a staff development programme in cooperation with the Academy of Management of the University of Groningen and Munthoff and Partners, a consultancy firm in the middle part of The Netherlands. The staff development programme is called: ‘The learning programme for senior policy directors of Municipality Emmen’. Twenty senior policy directors have undergone this intervention in 2008. A group of eighteen senior policy directors started in January 2009, and another group of twenty senior policy directors will start in September 2009.

The learning programme. The purpose of the programme was to create insights into a ‘process-directing role’ among the policy directors in addition to the ‘expert role’ they were

used to carry out. An important principle of the process-directing role is co-working with multiple actors, both internally and externally. To take part in the programme, employees within the Municipality Emmen could apply voluntarily for it by handing in a motivation letter to the HRM officer. Furthermore, an interview took place with a consultant of Munthoff and Partners. If the employee met the preconditions to take part, the employee was scheduled into the programme.

The programme consisted of six modules of two-day-lectures, assignments, and a list of literature. The six modules were spread over six months and the programme ended with an exam. The themes of the modules were: personal values and management drives within the work context; direction management and personal leadership; change management; force field analysis; process management, interaction and decision making; and lobbying in the future context of policy development. Reflection, the participants’ individual experiences, and theory were input for interaction during the entire course. The exam was twofold: a written part in which each participant individually exposed his or her point of view about policy development, and an oral part in which the participant advocated his or her written part in front of an exam committee consisting of the core lecturer1 _{(Munthoff and Partners), the} course manager2 _{(Academy of Management) and a professor (University of Groningen). The} (pro)positions had to be substantiated with the literature of the course. Nineteen out of the twenty participants (95%) in the learning programme completed the exam.

METHODS OF THE ROLE OF LEARNING IN CO-CREATING CHANGE Quasi-experimental research design

A number of evaluation designs can be applied to examine the outcome of interventions (Noe et al., 2003). When participants of the intervention can be expected to have similar levels of knowledge, behaviour, or results prior to the intervention, a post-test design is appropriate. This method can be strengthened by adding a comparison group, as similar to the other group as possible, who have not experienced the intervention; this helps to rule out alternative explanations for changes. In the case of the learning programme at the Municipality Emmen, we can assume that all participants of the intervention had similar levels of knowledge and behaviour prior to the intervention. Every participant had at least ten

years of work experience, of which at least two years at the Municipality Emmen. A precondition of the programme was that the participant had completed a high educational level in their previous education (HBO or WO level). At the time the participants started the programme they all had a policy related job within the Municipality Emmen. Therefore, to analyse the outcomes of the learning programme a quasi-experimental design is adopted (Gill and Johnson, 1997). The participants who took the programme in 2008 (N=20) functioned as the experimental group, and a comparison group (N=20) was composed by the HRM department of the Municipality Emmen. Eighteen people (90 %) of the experimental group, and nineteen people (95%) of the comparison group returned the survey.

Data collection

Brewer and Hunter (in Pennink, 2004) argue that, many social science research methodologies each have their own strengths and weaknesses. Instead of choosing one method, they argue that it is wiser to make use of combinations of methods3. According to the Durkheimian model, in which organisations are conceived of as social contexts for learning, all learning takes place inside individual human heads (Gherardi and Nicolini, 2003). This kind of learning involves insights and successful restructuring of organisational problems by individuals, which is reflected in the structural elements and outcomes of the organisation itself (Friedman, 2003). Therefore, organisational performance depends on the actions of its employees, and entails that organisational change can occur when the behaviour of these employees change. That is why the proposed causal relations in the context of co-creating change can be fruitfully examined at an individual level.

Collecting Data of Co-creation, New Knowledge, Interaction, Facilitation and Learning The post-test assessing the proposed hypotheses in this study was measured in two ways: an analytical survey (Gill and Johnson, 1997), and interviews to evaluate the results of the survey, to determine the effectiveness of the change.

Analytical survey. Outcomes in the form of a ‘how many’ and ‘how much’ line of enquiry – for example, ‘What were the outcomes of a particular managerial reorganisation?’ can likely favour the participants of the survey to identify (Yin, 1984). A survey can be readily designed to enumerate the ‘whats’. Appointments were planned for filling in the questionnaires. The questionnaires were handed over personally to prevent ‘non-response’ problems, and to explain and elaborate on possible problematic questions. Next, people were asked to fill in the surveys at a preferred time and place. The possibility was given to return the questionnaire anonymously, and it was explained that the questionnaires were processed confidentially. The filled in questionnaires were returned within a week by post or email (23%) and by handing in personally (77%).

Interviews. Because we accepted the importance of different research methods, and therefore chose for a pluralistic methodology, the results of the survey concerning the outcomes of the intervention were presented to participants of the learning course, to examine possible discrepancies between the outcomes of the surveys and constructs of social reality of the respondents. Three people from the group of participants of the learning programme functioned as informants in a focused one-to-one interview of approximately one hour (Emans, 2002; Baarda and De Goede, 1995; Yin, 1984). Before the interviews took place, the informants were informed about the context of the interview. It was explained that the interview concerned the results of the survey in which they had participated. Also, it was clarified that the interviews were confidential and that the answers would be published anonymously. The interviews were recorded with a tape recorder.

Measures

The structure of the questionnaires for both respondent groups – the participants of the learning programme and the employees who had not participated - were the same. Nevertheless, ‘facilitation’ was not measured among the comparison group because of its retrospective character. The entire questionnaire was in Dutch.

Measuring Co-creation, New Knowledge, Interaction, Facilitating, and Learning

‘facilitation’, ‘interaction’, ‘learning’, and ‘co-creation’, all these scales were developed based on the work of different theorists. Prior to testing the hypotheses, a rotated factor analysis was performed to confirm the assumptions about the amount of components in the developed scales needed to measure the variables. Based on the outcomes of these analyses, items in scales were removed or maintained (see Appendix B for the items in scales after principal component factor analyses, and Table 1 for characteristics of the scales).

TABLE 1

Characteristics of the scales

Scales It. M SD 1 Knowledge 4 .85 2.81 0.91 2 Interaction 11 .95 2.91 0.76 3 Facilitation 3 .65 3.75 0.56 4 Double-loop learning 4 .79 3.10 0.62 5 Triple-loop learning 6 .98 2.80 0.74 6 Crystallising 4 .90 3.15 0.86 7 Prototyping 3 .72 2.77 0.79 8 Performing 3 .64 3.09 0.52

analysis. The items inspired by Scharmer (2007) remained, and generated a coefficient reliability of .64 (with answer scale type 1, e.g.: ‘My way of working matches the image of how I think the policy development process should be’ ).

New knowledge, interaction and facilitation. ‘New knowledge’ was measured by a new scale, which was developed based on the definitions of tacit knowledge and codified knowledge by Weggeman (2000) (four items with answer scale type 2, e.g.: ‘Last year I gained new knowledge through other people’ s experiences concerning their work’ , and ‘Last year I gained new knowledge through studying theories and data concerning policy development’ ). The items forming the scale ‘interaction’ were derived from multiple sources. The items of this scale were developed based on existing literature, on research concerning listening (Scharmer, 2007; Senge et al., 2007) and conversation (Scharmer, 2007; Isaacs, 1999; Senge, 1992), and on an interview with an expert in the field of attention levels in interaction4. The scale was divided into two parts: 10 items concerning listening (answer scale type 2, e.g.: ‘During the past year the knowledge I gained surprised me’ , and ‘During the past year I looked at the policy development process from the perspective of my colleagues’ ), and 10 items concerning conversation (answer scale type 2, e.g.: ‘During the past year I avoided discussions to be nice’ , and ‘During the past year there was room to think and investigate aloud together with other people’ ). After conducting a principal component factor analysis, 11 out of 20 items remained. Striking was that ‘conversation’ and ‘listening’ were not two different components, as proposed. Based on the factor analysis, one variable of ‘interaction’ was formed. To assess ‘facilitation’ , four items were used from Metselaars’ (1997) DINAMO, which were adjusted to the learning programme (with answer scale type 1, e.g.: ‘The planned time schedule for the learning programme is realistic’ , and ‘The learning programme was there for me at the right moment’ ). The other three items were newly developed based on Tjepkema et al. (2002) (with answer scale type 1, e.g.: I was given the room to choose my own development direction within the learning programme’ ). From this scale three items remained after the rotated factor analysis. The Cronbach’ s alpha for the scales of ‘new knowledge’ , ‘interaction’ , and ‘facilitation’ were respectively .85, .95 and .65

Learning. To assess ‘double-loop learning’ and ‘triple-loop learning’ , sixteen items were formed including four items measuring ‘single-loop learning’ . The items concerning ‘single-loop learning’ were added to make an explicit distinction between the different levels of learning. These four items were removed before measurements. The twelve remaining

items were based on two newly developed scales taken from existing literature about learning (Scharmer, 2007; Agrysis and Schön, 1978; Pawlowsky, 2003), and an interview with a practitioner in life long learning5. Four items were concerned with ‘double-loop learning’ (with answer scale type 2, e.g.: ‘During the past year I reflected on underlying thoughts and assumptions in my acting’ , and ‘During the past year I drew conclusions about other people’ s underlying thoughts and assumptions in their acting’ ) with a Conbrach’ s Alpha of .80. From the scale of ‘triple-loop learning’ two items were removed based on the outcome of the rotated factor analysis (‘During the past year I reflected on underlying values and drives of my acting’ , and ‘During the past year I drew conclusions about underlying values and drives of my acting’ ). Six items remained (with answer scale 2, e.g.: ‘During the past year I reflected on other people’ s underlying values and drives of their acting’ , and ‘During the past year I drew conclusions about the organisation’ s underlying values and drives’ ), with a Conbrach’ s Alpha of .98.

RESULTS OF THE OUTCOMES OF THE LEARNING PROGRAMME Table 2 presents the results of the T-test on co-creation – divided in the stages ‘crystallising’ , ‘prototyping’ , and ‘performing’ – of the respondents who had undergone the intervention, and of the group of respondents who had not. The T-test shows that there is a significant difference in scores between the two groups on ‘crystallising’ (t=5.51, p < .01), ‘prototyping’ (t=2.63, p < .01), and ‘performing’ (t=2.26, p < .05). These results of the survey support that participants of the learning programme scored higher on co-creating change, in the sense of our definition of co-creating, than the employees who did not participate in the learning programme.

TABLE 2

Results of Scores on Co-creation of Participants and Comparison Group

Variable Group N M SD t-value df

1 Age in years Participants 18 47.56 9.91 .25 29.56 Comparison 19 46.84 7.71

2 Years of service Participants 18 17.33 9.25 .11 35.00 Comparison 19 17.00 8.62 3 Crystallising Participants 18 3.75 .45 5.51 ** 35.00 Comparison 19 2.61 .77 4 Prototyping Participants 18 3.09 .65 2.63 ** 35.00 Comparison 19 2.46 .80 5 Performing Participants 18 3.28 .51 2.26 * 35.00 Comparison 19 2.91 .47 * p < .05 (1-tailed) ** p < .01(1-tailed)

TABLE 3

View of informants (3 participants) on the results on co-creation of change of the group who participated and of the group who did not participate in the learning programme Crystallising

There are differences in the images on policy directing between the two groups. The participants of the learning programme gained new views and insights, which form principles to create an image of policy directing. The respondents of the comparison group do not have this information and will form their image based on ‘old’ views and insights or do not have a clear image of policy directing at all.

Informant 1

I am more aware of how I create the process. I am directing the process more consciously. When somebody is more aware of the possibilities of how to direct the policy process that will create a clearer image of the policy directing process.

Informant 2

For the experimental group, the world of policy directing has become more complex. The participants of the learning programme gained different insights and views on policy directing. This creates confusion and makes it harder to create a clear image of how to direct a policy process. The comparison group do not have all that new input, so they refer to the way they are used to do it for years. That seems to me to be easier, and I think it will create a clearer image how to direct a policy process.

Informant 3

Prototyping

The gained knowledge and skills, plus the appreciation of the organisation to the participants by giving them the opportunity to participate in the learning programme, created confidence in trying a new way of working among the participants.

Informant 1

The participants of the learning group are experimenting with a new way of working. When looking at my self, I recognise that. Now I am in a process that I am doing things I have never done before. My direct colleagues are surprised by that. They are my frame of reference.

Informant 2

The comparison group has less input for experimenting. Therefore, for me it is logical the respondents of the comparison group do not or at least to a lesser extent experiment with a new way of working.

Informant 2

Performing

Performing is measured by two different quantities. The comparison group has another

frame of reference than the experimental group. I am sure about that. Informant 1 My way of working has definitely been influenced by the learning programme. My direct

colleagues consolidated that. Informant 2

Despite that I notice that there is a little difference in performing between the comparison group and the experimental group, the difference between the two groups is not very large. It takes time to transmit theory into practise. That could be a reason that scores of performing among the participants are not high yet.

Informant 3

experimenting with that new way of working (prototyping), are still in the process of institutionalising the crystallised image, or have not even reached that stage (performing).

RESULTS OF THE ROLE OF LEARNING ON CO-CREATING CHANGE The mediated effects of learning on co-creating change

Table 4 presents descriptive statistics about the variables and correlations among the variables used in the regression analysis. The mean of ‘facilitation’ scores is relatively high compared to the means of ‘new knowledge’ and ‘interaction’ . The means for ‘double-loop learning’ and ‘triple-loop learning’ exhibit a difference of .30, which is relatively small. The means of ‘crystallising’ , ‘prototyping’ , and ‘performing’ are inconspicuous when compared to each score. It is striking that the standard deviation of ‘performing’ is relatively low in comparison to the standard deviation of ‘crystallising’ and ‘prototyping’ .

Bouma and Emans (2005) explain that, to assume a mediating relationship four conditions have to be met: A direct relationship has to exist between the independent variable (new knowledge, interaction, facilitation) and the dependent variable (crystallising, prototyping, and performing); the mediators (double-loop learning and triple-loop learning) need to have a positive relation with the independent variables; a positive relation has to exist between the mediators and the dependent variables; and the significant relation between the independent variables and the dependent variables has to disappear, or at least decrease when controlled by the mediators.

TABLE 4

Characteristics of Variables and Pearson Correlations a

Variables Item N M SD 2 3 4 5 6 7 8 1 Knowledge 4 .85 2.81 0.91 .94 ** .02 .35 * .61 ** .71 ** .56 ** .42 ** 2 Interaction 11 .95 2.91 0.76 .37 .32 * .60 ** .70 ** .52 ** .41 ** 3 Facilitation 3 .65 3.75 0.56 .42 * .46 * .41 * .47 * -.10 4 Double-loop learning 4 .79 3.10 0.62 .60 ** .16 .31 * -.32 5 Triple-loop learning 6 .98 2.80 0.74 .53 ** .40 ** .29 * 6 Crystallising 4 .90 3.15 0.86 .42 ** .23 7 Prototyping 3 .72 2.77 0.79 .25 8 Performing 3 .64 3.09 0.52

a _{n = 37 (knowledge, performing), n = 36 (interaction, double-loop learning, triple-loop learning, crystallising) n = 17 (facilitation)}

Although the relations of the independent variables with the mediators are all positive, the relations between ‘triple-loop learning’ and the independent variables correlate stronger than the relations between ‘double-loop learning’ and the independent variables. However, the correlations exhibit that the second condition has been met. When examining the third condition, the correlations in table 4 exhibit a partial support for hypothesis 4 - a high level of ‘double-loop learning’ and/or ‘triple-loop learning’ leads to high co-creation of change. ‘Double-loop learning’ has no significant correlation with ‘crystallising’ , a significant medium positive relation to ‘prototyping’ , and no significant negative relation with ‘performing’ . Because of its non-significance and because of its negative correlation to ‘performing’ , it is not taken into the regression analyses. In contrast, ‘triple-loop learning’ has significant positive relations with all three variables of co-creating change. The results of the two stage regression analyses of ‘triple-loop learning’ on the dependent variables (table 5) exhibit a significant linear relation between ‘triple-loop learning’ and ‘crystallising’ , ‘prototyping’ , and ‘performing’ (β= .54, p <.01; β= .40, p < .05 and β= .32, p < .05).

TABLE 5

Results Regression of Triple-loop Learning on Crystallising, Prototyping, and Performing

Crystallising Prototyping Performing

Step 1 2 1 2 1 2

Scale Beta Beta Beta Beta Beta Beta

1 Age -.05 -.01 -.09 -.06 .17 .20

Years of service .03 .09 -.01 .03 .01 .06

2 Triple-loop learning .54 ** .40 * .32 *

R2 _{.00 .29 * .01 .17 .03 .14}

The standardised regression coefficients are presented. N =36

* p<.05

** p<.01

To test the mediated regression effects, regression analyses were conducted for each dependent variable. In the first stage, the dependent variable (for instance ‘crystallising’ in the first part of table 6) was regressed on the background variables age and years of service to control the effects. Then an independent variable (for instance knowledge in table 6) was added in the second stage. Finally, ‘learning’ was added to test the mediated effect on the relationship between the dependent variable and the independent variable.

analyses (β= .71, p <.01; β= .56, p < .01 and β= .39, p < .05) show a direct relation with the dependent variables, ‘new knowledge’ explains no significant share of the variance in ‘performing’ (R2 =.19, ns). However, ‘new knowledge’ does explain a significant share of the variance in ‘crystallising’ and ‘prototyping’ (R2 =.50, p<.01; R2 =.32, p<.01). These results support acceptance of hypothesis 1.1 and 1.2: Acquiring ‘new knowledge’ regarding the context of the change will lead to high ‘crystallising’ and high ‘prototyping’ . Finally, in the third steps of the regression analyses, the significant relations between ‘new knowledge’ and ‘crystallising’ , and ‘new knowledge’ and ‘prototyping’ have decreased a little when controlled for ‘triple-loop learning’ . However, ‘triple-loop learning’ shows no significant beta. Therefore hypothesis 1* - the relation between acquiring ‘new knowledge’ and co-creating change is mediated by ‘triple-loop learning’ - cannot be accepted.

Interaction. Table 7 presents the results of regression analyses for ‘interaction’ , controlled for the control variables. The steps 2 exhibit a direct relation between ‘interaction’ and ‘crystallising’ , ‘prototyping’ , and ‘performing’ (β= .72, p <.01; β= .52, p < .01 and β= .44, p < .01). ‘Interaction’ does explain a significant variance of 51% (p<.01), 27% (p<.05) and 22% (p<.05) in respectively ‘crystallising’ , ‘prototyping’ , and ‘performing’ . Therefore, hypothesis 2 can be accepted; for every dependent variable –which together represent the three stages of co-creation- a positive linear relation exists with ‘interaction’ . Steps 3 show that the relation between ‘interaction’ and ‘crystallising’ and between ‘interaction’ and ‘prototyping’ decreases when controlled for the mediator ‘triple-loop learning’ . However, ‘triple-loop learning’ shows no significant beta. The relation between ‘interaction’ and ‘performing’ decreased, but also not significantly. Therefore hypothesis 2* cannot be accepted. Our results show no support for the proposed mediated role of ‘triple-loop learning’ in the relation between ‘interaction’ and co-creation of change.

TABLE 8

Results Regression Analyses of Facilitation on Crystallising, Prototyping, and Performing with Mediator

Crystallising Prototyping Performing

Step 1 2 3 1 2 3 1 2 3

Scale Beta Beta Beta Beta Beta Beta Beta Beta Beta

1 Age -.31 -.18 -.22 -.09 .27 .24 .18 .16 .16 Years of service -.04 -.12 .02 -.14 -.26 -.12 .01 .02 .03 2 Facilitation .38 .16 .56 * .34 -.06 -.07 3 Triple-loop learning .43 .42 .03 R2 _{.00 .23 .36 .02 .30 .42 .03 .04 .04}

‘facilitation’ and ‘prototyping’ disappears when controlled for ‘triple-loop learning’ . However, as is the case with ‘new knowledge’ and ‘interaction’ , ‘triple-loop learning’ shows no significant beta. Therefore, hypothesis 3* cannot be accepted. The results do not support a mediating role of ‘triple-loop learning’ in the relation between ‘facilitation’ and co-creating change.

TABLE 9

Results Regression Analysis of Knowledge, Interaction, and facilitation on Crystallising, Prototyping, and Performing

Crystallising Prototyping Performing

Step 1 2 1 2 1 2

Scale Beta Beta Beta Beta Beta Beta

1 Age -.31 -.08 -.09 -.24 .18 .16 Years of service .04 .41 -.14 -.07 .01 -.07 2 Knowledge .39 .45 -.45 Interaction -.69 -.01 .15 Facilitation .68 .52 -.10 R2 _{.10 .37 .02 .47 .03 .16}

The standardised regression coefficients are presented. N =17

* p<.05

** p<.01

Table 9 presents the results of the regression analyses for ‘new knowledge’ , ‘interaction’ , and ‘facilitation’ together on the dependent variables. None of the independent variables has a significant relation with co-creation of change when all three are taken together. Therefore, we conclude that acquiring ‘new knowledge’ , ‘interaction’ , and ‘facilitation’ have no joint contribution to co-creation of change, but only independently in the sense of the results of the regression-analyses in tables 6, 7 and 8.

co-creating change, we can conclude that ‘triple-loop learning’ also has strong correlations with all three stages of co-creating change: ‘Crystallising’ , ‘prototyping’ , and ‘performing’ .

TABLE 10

Overview of Acceptance of Hypothesised Relations based on Our Results

Hypothesis Accepted/

Not Accepted

H 1.1. Acquiring ‘new knowledge’ regarding the context of the change will

lead to high ‘crystallising’ . Accepted

H 1.2. Acquiring ‘new knowledge’ regarding the context of the change will

lead to high ‘prototyping’ . Accepted

H 1.3. Acquiring ‘new knowledge’ regarding the context of the change will

lead to high ‘performing’ . NA

H 1.* The influence of acquiring ‘new knowledge’ regarding the context

of the change is mediated by ‘learning’ . NA

H 2.1. A high level of ‘interaction’ will lead to high ‘crystallising’ . Accepted

H 2.2. A high level of ‘interaction’ will lead to high ‘prototyping’ . Accepted

H 2.3. A high level of ‘interaction’ will lead to high ‘performing’ . Accepted

H 2.* The influence of ‘interaction’ on co-creation is mediated by

‘learning’ . NA

H 3.1. A high level of ‘facilitation’ will lead to high ‘crystallising’ . NA

H 3.2. A high level of ‘facilitation’ will lead to high ‘prototyping’ . Accepted

H 3.3. A high level of ‘facilitation’ will lead to high ‘performing’ . NA

H 3.* The influence of ‘facilitation’ on co-creation is mediated by

‘learning’ . NA

Interaction effects of learning on co-creation

between the concerning independent variable and ‘triple-loop learning’ to test interaction between ‘triple-loop learning’ and the independent variables.

The nine conducted regression analyses yielded two values that support an interaction effect: interaction between ‘new knowledge’ and ‘triple-loop learning’ ( R2= .32, p < .01), and between the variables ‘interaction’ and ‘triple-loop learning’ ( R2_{= .24, p < .05). Both} interaction effects were found only for ‘prototyping’ . Figures 3 and 4 illustrate the interactions.

FIGURE 3

Prototyping: Interaction between Interaction and Triple-loop Learning 1,5 1,7 1,9 2,1 2,3 2,5 2,7 2,9 3,1 3,3 3,5 Low High Interaction Pr ot ot yp in g

Triple loop learning low Triple loop learning high

FIGURE 4

Prototyping: Interaction between New Knowledge and Triple-loop Learning 1,5 1,7 1,9 2,1 2,3 2,5 2,7 2,9 3,1 3,3 3,5 Low High New Knowledge Pr ot ot yp in g

We found that there was a low effect of ‘new knowledge’ on ‘prototyping’ when ‘triple-loop learning’ was low. In contrast, a strong effect appeared on the relationship between ‘new knowledge’ and ‘prototyping’ when ‘triple-loop learning’ was high. High ‘triple-loop learning’ appeared to facilitate the impact of ‘new knowledge’ on ‘prototyping’ . We also found that ‘interaction’ had a positive relation with ‘prototyping’ when ‘triple-loop learning’ was both low and high, yet when ‘triple-loop learning’ was high the effect was stronger. High ‘triple-loop learning’ also appeared to facilitate the impact of ‘interaction’ on ‘prototyping’ .

Therefore, we conclude that our findings support that ‘triple-loop learning’ has a moderating role in the relation of ‘new knowledge’ and ‘prototyping’ , and in the relation of ‘interaction’ and ‘prototyping’ .

DISCUSSION ABOUT THE PREDICTORS AND THE EFFECT OF LEARNING ON CO-CREATING CHANGE

Discussing the results of the role of learning on co-creating

In order to succeed in a turbulent and complex business environment as an organisation, the view on organisational change has been expanded enormously. In this study it is believed that, to succeed in a knowledge and innovation driven economy value creation can be increased through co-creating. Nevertheless, proposed outcomes of change interventions have often not been met, and therefore change agents need insights into what change interventions can be used successfully. In this study, we assumed that the role of a deep level of learning during change interventions would have a remarkable influence on achieving co-creating change. Furthermore, we expected that new knowledge, interaction and facilitation were the predictors of deeper levels of learning and of the co-creation of change. These assumptions were based on three existential stances on the human condition: knowing, action, and feeling. This study offers the first empiric research into a deep level of learning as a change intervention in relation to new knowledge, interaction, and facilitation in order to co-create sustainable. The results point to interesting and novel findings.

of interaction, referred to as dialogue and shifts of attention in listening, seems to lead to change behaviour (Scharmer, 2007). This implicates that, to co-create change it is important that employees interact with each other in the form of a ‘collective space’ (see Scharmer, 2007), wherein they develop a (new) collective image about the purpose of the change (see Pennink, 2004). Gathering new theories and sharing experience, thoughts, and ideas among employees, forms new images that predict co-creating change. Contradicting the beliefs of various change management practitioners (Burnes, 2004; Noe et al., 2003; Tjepkema et al., 2002) that facilitation is vital for successful behavioural renewal, the results show no support for facilitation as a strong predictor for co-creating change. Facilitation only seems to be a significant predictor for prototyping, but not for crystallising and performing. This is striking, since crystallising, prototyping, and performing are all seen as stages of co-creating change, whereas prototyping is seen as the second stage (Scharmer, 2007). This implicates that, to achieve experimenting among employees during a co-creating change process time management, room for taking one’ s own responsibility, and confidence in and support of superiors need attention from the change agent. Thus, facilitation gives rise to experimenting with a new way of working in a manner that employees actively try to do things in order to institutionalise the crystallised image. Nevertheless, our results implicate that facilitation is less important to achieve the stage of crystallising that image and to achieve the stage of performing. That facilitation - such as defined in this study - is important for reaching experimenting among employees is in line with research about the impact of being trusted (trust in subordinate by superior) on the subordinate’ s behaviour (Brower et al., 2009). They found support for the fact that managers’ trust in the subordinates should lead to high-quality interactions that convey a sense of empowerment and confidence to the subordinate. Consequently, subordinates are motivated more to exert greater effort within and beyond their prescribed roles. In other words, when employees experience room for taking their own responsibility, support, and confidence in their superiors they are motivated more to experiment with a new way of working. They argue that experimenting leads to better performance by subordinates. That our results do not support this - that facilitation leads to high performing - may be explained in part by that this last stage of co-creation is not reached (yet) among the experimental group who had undergone the learning programme (see interviews with informants).

interaction, and facilitation with the three dependent variables altough the results show that all relations between the predictors and ‘crystallising’ and ‘prototyping’ decreased, when controlled for the mediator ‘triple-loop’ learning.

Nevertheless, an interaction effect was found, and two significant patterns of moderated results were depicted. The interactions with ‘prototyping’ suggest that ‘triple-loop learning’ amplifies the effect of ‘new knowledge’ and ‘interaction’ in their relations with ‘prototyping’ . When ‘triple-loop learning’ is high, ‘new knowledge’ and ‘interaction’ both exhibit greater and positive relationships with ‘prototyping’ . This suggests that acquiring ‘new knowledge’ , and providing conversations and shifting attention in listening had more effect on experimenting when the participants of the learning programme learned on the level of ‘triple-loop learning’ . This is in line with the belief that collective reflection on how to govern organisational and individual values and drives gives rise to co-creating change. By concluding that ‘triple-loop learning’ , which is a level of learning pinpointed as an emotional experience, has a moderating role in the sense that when ‘triple-loop learning’ increases the relations between ‘new knowledge’ and ‘prototyping’ , and between ‘interaction’ and ‘prototyping’ becomes stronger, the presence of the positive role of emotional effectiveness during learning in order to co-create change is confirmed. This is in line with Pawlowsky’ s (2003) statement about organisational learning that knowledge, feelings, and action must all three be considered complementary if organisational learning is to be promoted.

Concluding, our results indicate that the relation between the components new knowledge, feelings in sense of facilitation, and action as part of interaction is likely to be an important aspect of change. People in organisations have to understand or create new images, and have to feel that it is right to adopt new views and that they are able to act on those accordingly. The sharing of knowledge or of assumptions, which is necessary for true conversation and listening, does not have a chance if people are afraid to lose in such a process. Therefore, learning as a change intervention makes it incredibly important to establish, create, and nurture a culture of trust.

Theoretical and practical implications of learning as a change intervention

themselves who have to change their way of working. Therefore, the information must be surprising, so people really hear new things, and the stories and information must be inspiring, so people feel they want, can, or even have to change. To surprise and inspire people during a change intervention, it demands of the change agent to act expertly in the sense that he or she knows what is going on in the particular business, organisation and/ or teams. Furthermore, it demands creativity and renewal in approaching the people to ‘tempt’ them to deeply (re)think the information in order to let valuable and useful knowledge emerge.

To reach a high interaction level among the participants of an intervention, room must be created to allow them to shape their own viewpoints and assumptions (more) explicitly. People have to be transparent towards each other about what they experience and where they stand, how they work, what they feel and what they want. Preceding the ability to make one’ s assumptions and points of view explicit, participants must search for ‘why do I, or we want that?’ as an individual, but also at the group and organisational level. The role of the change agent is to facilitate this and to trigger the participants to start with the search for these underlying values. During the process it is important to keep the purpose of the intervention in scope. One person will look for the causal relations and the underlying patterns; the other will look for a connecting thread, the history or evidence. The focus is not to live through undesirable circumstances, but to let the future emerge (Scharmer, 2007). When learning to see the different points of view, people hear (listen) and speak (conversation) about different realities. When doing that in an unusual way, people see a new image emerge to co-create change. It is about creating situations wherein people get grip on the situation together without weakening another for the sake of bettering oneself. Open and honest conversations without shielding the others from negative feelings are the way to show equivalence among participants, which brings listening and conversation to a higher level. Therefore, it is incredibly important that the change agent recognises defensive attitudes immediately, and not afterwards, and creates an atmosphere of mutual trust among the participants of the intervention.

motivate them, but you cannot force them to participate in (re)thinking and learning. Furthermore, as argued above management of learning makes it necessary to understand learning not only as a matter of cognitive learning, but also as a matter of values, emotions, and behaviour. Therefore, the participants must experience room for taking their own responsibility, and confidence in and support of superiors to let these values, emotions, and behaviour become evident to co-create change. Superiors must create an atmosphere where mistakes are allowed, and experimenting with a new way of working is promoted. Besides that, it is important for employees to experience the confidence of the superior for experimenting, the employees must also have the confidence that they can do it. Thinking and doing must be linked in order to create the proposed change. Therefore, employees must have the right competences, and need to know how to properly use them. The change agent can facilitate that by, for example, supplying the right tools, allocating resources, or providing trainings.

Limitations and future research

Sample. Even though a comparison group was included in this study, all respondents were from the same organisation. Furthermore, this study is limited by the lack of a wider sample, which would achieve a more cross-sectional representation of the context. It may be precarious to conclude in terms of causal relationships when all data comes from one source and the sample is small (Galema, 2008). In order to generalise our findings, future research should be conducted in other sectors as well, because of the specific character of the public sector, and with a wider sample.

Data collection. The data was collected personally. Therefore, the response could be higher than normal. Even though we verified the results of the surveys by conducting interviews, we recognise that some positive bias could have been taking place, because respondents might tend to present themselves or the learning programme more positively than in fact they were. In future research this bias could be prevented by including more peers, superiors, or even customers. It would also be interesting to include the role of the change agent. The role of the change agent could have had an important influence on how the learning programme was perceived (Bennebore Gravenhorst et al., 2003).