GPSA Note 12 ‘LEARNING JOURNEYS’ FOR ADAPTIVE MANAGEMENT-WHERE DOES IT TAKE US? 1 Learning-by-doing and adaptation happens all the time at the frontline, and as a means of survival. So how can it be reflected across the sector as a whole – from individual citizen through to international donor through adaptive management? March 2016 INTRODUCTION In development initiatives, learning happens all grate it into established institutional frameworks. the time through continuous information inputs, Consequently, it comes as no surprise that interaction, action, assessment and adaptation ‘learning for development’ could be seen as a to solve complex problems. Even so, traditional rather insular or introspective exercise, relevant project cycle management and accompanying only to a small number of academics or devel- knowledge initiatives often compartmentalize opment practitioners, instead of being a driver such learning into ‘lessons’ or ‘success stories’ of social change towards increased impact. at the end, or do not sufficiently take into ac- count the various learning needs among the So how could we frame the ‘learning agenda’ wide spectrum of actors involved. For instance, going forward so that continuous, learning- while clients and practitioners on the ground oriented adaptation is inclusive and seen as a may hope to influence the sector as a whole to driver towards large-scale impact? And how do development differently – a concept which does this challenge affect existing knowledge has recently gained momentum2 – uptake management practices, and what have we of projectized learning tends to be difficult in learned to date? large funding institutions on the grounds that it is too specific or small-scale. Conversely, re- Looking across the existing and growing body strictive donor behavior can hamper learning in of literature both in ‘systems thinking’, be- operations. And citizens may be left out of the havioral theory and adaptive management learning process altogether once information practices for development, this GPSA Brief- or feedback has been ‘extracted’. Knowledge ing Note seeks to clarify some of the underly- communities of different kinds (communities ing concepts and how they are and/or could of practice) seek to break compartmentalized be applied in practice. It also suggests some learning silos, but may lack themandate to inte- additional practical steps, going forward. 1 This GPSA Learning Note was written by Charlotte Ørnemark, GPSA Knowledge & Learning Team, World Bank. February, 2015. 2 See: www.doingdevelopmentdifferently.com To access the links in this note, go to http://gpsaknowledge.org /1 I. WHY IS ADAPTIVE MANAGEMENT RELEVANT TO SOCIAL ACCOUNTABILITY? GLOSSARY In social accountability, information or feedback from different stakeholder groups are used as Lost in translation? This is what triggers to improve services and set up engage- the dictionary says… ment mechanisms where citizens can hold ser- vice providers or decision-makers to account. Knowledge: Facts, information, and skills acquired through expe- With a multitude of interests and incentives at rience or education; the theoreti- stake, it is a very unpredictable process, even cal or practical understanding of when the end goals are shared between citi- a subject. zens, civil society and government actors. For Knowledge management: Effi- it to work, all parties will have to act and learn, cient handling of information and trying alternative paths towards joint problem- information resources within an solving. organization or system. Learning: The acquisition of knowledge or skills through study, experience, or being taught. “Behind every policy is an assumption about human behavior. (…) Sometimes the assump- Adapt: Make (something) suitable for a new use or purpose; modify. tions can be wrong.” Loop: A structure, series, or World Development Report 2015 on “Mind and Society” process, the end of which is con- nected to the beginning Feedback loop: The modification There are many reasons why change hap- or control of a process or system pens. To adapt, however, we need information by its results or effects, for exam- ple in a behavioral response. or experiences that make us question our ex- isting behavior. Yet all information is screened Tacit knowledge: Unwritten, through mental models shaped by many socio- unspoken, and hidden vast storehouse of knowledge held cultural factors and past experience. Predicting by practically every normal hu- behavioral response is therefore nearly impos- man being, based on his or her sible. It has to be continuously tested. This is emotions, experiences, insights, true among our own peers at an individual lev- intuition, observations and inter- el, and even more so when we seek to influence nalized information. change in contexts different to our own where Explicit knowledge: Articulated the response of one actor in an interdepend- knowledge, which is expressed and recorded as words, numbers, ent system also influences that of another. This, codes, mathematical and scien- among other things, was discussed in the World tific formulae etc., and which is Development Report (WDR) 2015, focusing on easy to communicate to others ‘How a better Understanding of Human Behav- via books, on the web, and other ior Can Improve Development Policy’. It outlines visual and oral means. different “frames” through which we see the http://www.oxforddictionaries.com/ world, and how we fill in missing information http://www.businessdictionary.com based on default assumptions rather than on questioning our mental models and look for evidence. /2 Relevant information (in the form of timely and to program inefficiencies or larger-scale failure. relevant ‘feedback’ and data) can, according to the same behavioral theorists, help to break Contrary to short loops of experimentation through existing mental models and open up and learning from “failure” that can lead to in- for new response mechanisms and forms of cremental adaptation and efficiency improve- collaboration. This will in turn lead to learning- ments, large-scale failure can be devastating in- by-doing and move issues forward towards stitutionally and often lead to sanctions in terms more systemic change. of cut funding. It’s ‘learning too late’. The issue, then, is how to accommodate Just like specific interventions or projects need learning needs by making better use of vari- to be regularly assessed, monitored and re- ous efforts to open up governance, generate assessed, organizations are also constantly channels for timely knowledge exchange, and transforming entities that adapt and change ensure citizen engagement in the co-creation depending on the individuals inside them (their of solutions that lead to social accountability. skills, capacities and learning), the policies and The focus on local, iterative problem-solving processes that guide them, and their interac- has inspired efforts such as the Doing Devel- tions with their operating environment. Most opment Differently movement to look at how such systems of human interaction are ‘nested’ we learn through problem-driven iterative pro- or placed in larger systems of e.g. public sec- cesses in development3 . Others have looked at tor reform or social change. The way one ac- complexity theory and have translated systems tor in such a system behaves or interacts with thinking into practical efforts to learn and adapt others can potentially affect overall system per- alongside other actors in order to increase the formance. In other words, incremental changes effectiveness of their interventions4. and learning has to happen at multiple levels si- multaneously for a more systemic shift forwards, It has also been argued that processes where towards a desirable development outcome. a lot is already known about cause and ef- fect (e.g. ‘people who sleep under a bed net to avoid mosquito bites are less likely to get II. WHY IS ADAPTIVE MANAGEMENT A TIMELY TOPIC? malaria, therefore we distribute bet nets to families’) may be better suited to a more lin- Combining or complementing increasingly ear implementation model than those where open governance processes for participation many different implementation paths are pos- and data generation with cyclical citizen-driven sible, and where there is a higher degree of feedback opens up new avenues for learning interdependency of actors to achieve a spe- and adaptation. It also departs from citizen cific objective. Yet, even so-called ‘predictable’ feedback as being a largely ‘extractive’ exer- interventions may backfire due to a lack of cise where information goes primarily one way contextual or social variables being taken ac- (from user to provider) based on perceptions count, as we all know (e.g. ‘distributed bed nets alone, rather than engaging in an increasingly were sold by the poorest’ or ‘women getting informed dialogue on alternatives for action. up from bed before sunrise for food prepa- In processes of complex change with many ac- ration and get exposed to mosquitos then’). tors (as opposed to those with high causal pre- Without the means to engage in learning-ori- dictability), the ‘ability to act on information’ ented monitoring along the way, it would lead 3 Andrews, M., Pritchet L., Woolcock, M., Escaping Capability Traps through Problem Driven Iterative Adaptation (PDIA), Harvard University, June 2012 4 Ramalingam, B., Aid on the Edge of Chaos: Rethinking International Cooperation in a Complex World, Oct., 2013 /3 of all in the system then becomes important. A more learning-or action-oriented approach Without action, no learning. What is ‘action- can be seen in e.g. the GPSA funded project able’ may differ between different interdepend- in the Philippines targeting a sub-set of benefi- ent actors across the system, and will certainly ciaries of the national conditional cash transfer differ from an independent citizen to a large in- program, the Pantawid Pamilyang Pilipino Pro- ternational development agency. How informa- gram (4Ps).Existing mandatory information ses- tion is used again affect overall system perfor- sions with the poorest are being used to both mance. This goes into a much more nuanced, teach them about their rights and help them and often political, understanding of what types try these concepts out in real life by monitor- of knowledge inputs different actors may need, ing their own children’s health and education and when, in order to take a change agenda attendance and engaging with local public of- forward. ficials on local service delivery needs. But even with this innovative practice in place, the actual For instance, being asked to give feedback on learning and change in behavior of cash trans- a service may in itself be an empowering ac- fer beneficiaries needs to be monitored from tion for some citizens that could spur further a ‘learning’ and ‘action’ perspective in addition engagement, but may not necessarily lead – in to fully comprehend how this approach lead to itself – towards any paths of alternative action learning and changed behavior of beneficiaries. or incremental behavior change at an individual This would be in addition to the more formal level. If it does, the ‘action potential’ may differ compliance monitoring that the government depending on whether they are men or wom- and external funders are interested in. en, middle class or marginalized etc. Beneficiaries from the conditional cash transfer (CCT) program implemented by the Philippines Department for Social Welfare review their self-evaluation forms about their children’s educational and health status. Involving people in both self- and third party monitoring of local services is an innovation of the GP- SA-funded project aimed at improving the integrity and compliance of the CCT. (Photo: C. Ørnemark, GPSA/World Bank). /4 III. WHAT ARE THE ORIGINS OF ADAPTIVE MANAGEMENT AND HOW ARE CON- CEPTS BEING USED IN DEVELOPMENT? The concept of adaptive management refers to a Some have defined adaptive management as process that promotes flexible decision-making an approach where “management is treated as that can be adjusted in the face of uncertainties a deliberate experiment for the purpose of learn- as outcomes from actions and other events be- ing” 6 , whereas others put emphasis on its utility come better understood. This requires careful, for exploring alternative ways to meet objec- learning-oriented, monitoring and the system- tives, and adapting project or process man- atic incorporation of stakeholder feedback as agement practices based on a combination of an integral part of the overall management of knowledge and ‘real-time’ learning-by-doing. It an organization’s or system’s operations. Adap- is also aimed at making decision-making more tive management underpins an operational inclusive at all levels, drawing extensively on definition of learning as a relatively permanent two-way sharing of information, and getting change in behavior potential, resulting from ex- feedback from stakeholders in an iterative man- perience. This is in contrast to ‘latent learning’ ner.7 which refers to learning that is not necessarily reflected in overt behavior change. In other words, adaptive management as it The concept was first developed and applied was originally conceived, was not designed to in the natural resources sector in the late 1970’s find out ‘what works best’ to solve a particular and 1980’s5. It recognizes the high degree of development problem, but also to holistically variability and interdependency in ecological understand system behavior in relation to the systems, and the need to apply a ‘systems ap- problem being addressed. proach’ and close monitoring of any external intervention in a system to understand better An implication for social accountability initia- how existing holding patterns (or ‘equilibrium’) tives could be to place specific public sector will be affected and/or lead to change in overall problem-solving interventions in a broader systems performance. (Just like the introduction systemic perspective to ensure that feedback is or removal of a species can disrupt an ecosys- inclusive of those who are currently marginal- tem equilibrium). ized and/or who have a vested interest to resist change to better understand systemic hold- Lately, this concept has gained traction also ing patterns (including gender inequalities and in other areas of development particularly marginalization). in highly unpredictable processes of change where power and politics influence existing Another implication – and one which the GPSA governance deficiencies. Using systems analy- has sought to at least partially address through sis and monitoring over time, existing ‘holding its knowledge and learning component includ- patterns’ that hinder system effectiveness can ing the GPSA knowledge platform – is to en- be revealed and observed, and usually goes gage in a continuous dialogue both with peers beyond the life-span of an individual project. and funders about lessons on a more continu- ous basis, bridging practitioner learning with 5 Holling, C. S. 1978. Adaptive Environmental Assessment and Management. Chichester, UK: John Wiley and Sons., and Walters, C. J. 1986. Adaptive Management of Renewable Resources. Macmillan, New York. 6 See: www.greenfacts.org 7 U.S. Department of the Interior /5 funding agencies’ internal learning systems and ture highlighting the need to “tap people’s com- incentives. Knowing how ‘learning journeys’ mitment and capacity to learn at every level in happen in practice could also help to focus a the company”, drawing on dimensions such as more substantive dialogue around how to do (i) systems thinking, (ii) personal mastery, (iii) development differently through a more adap- mental models, (iv) building shared vision, and tive management approach as a means towards (v) team learning. It goes back to the notion that achieving joint objectives (see e.g. a recent it needs more than individual or project-specific GPSA World Bank Brown Bag Lunch discus- course correction to have a transformational sion9 on the subject, featuring guest speakers impact on the whole organization or system or from the Overseas Development Institute). sector. Practically, it helps to think of different ‘learning systems’ and who would need to be part of such a system in order for more deep- Other related concepts rooted changes in values or discourse to occur.8 Often used as shorthand for learning for adap- tive management, ‘adaptive learning’ generally refers to organizational learning that focuses on ”Organizations are either learning or they past successes and failures to make incremental are dying. (…) Learning is a process that improvements to their offerings in response to enhances knowledge, and knowledge is the their changing environments. It is easy to get capacity for effective action.” lost in semantics between those who argue that Peter Senge, on what it will take for the World Bank to adaptive learning implies continuous ‘coping’ always evolve through learning, from “The See-Saw of with the environment through a multitude of World Bank Learning”, J. Haynes, IEG, Aug. 25, 2015 smaller scale course corrections (without neces- sarily questioning underlying beliefs and regu- latory frameworks), and those who believe that Finally, the concept of ‘feedback’ and ‘feedback incremental course correction constitutes only loops’, useful as it is to illustrate the iterative na- one step towards more deep-going learning ture of interactions between service provider leading to transformational change. and service user, can also cause some confu- sion. Some disciplines10 refer to it as a means In the logic of ‘transformational change’, indi- to give citizens ‘voice’ in decision-making that viduals or organizations may course correct, affects them as both a means and end in itself. but such course correction will only lead to sys- Others – such as in behavioral economics, epi- tem transformation if they go through several demiology or psychology – refer to it as a spe- loops of scrutiny where external feedback com- cific response mechanism based on a behavio- plements other evidence and knowledge to ral response or action at different levels. Finally, repeatedly challenge the dominant discourse, in monitoring and evaluation, it is more com- mental models, and key assumption (see figure monly used as a means to an end – namely to 1, Section 5). enhance internal learning through regular ‘real- ity checks’ as a part of the ongoing monitoring Pioneered by Peter Senge in the 1990s, and still and learning process. The below box explains used today, is the organizational learning litera- further. 8 Peter Senge. The Fifth Discipline. 1990 9 http://www.thegpsa.org/sa/event/bbl-doing-development-differently-politically-smart-and-adaptive-approaches-address-governance 10 E.g. in relation to public sector or administration reform. /6 BOX 1 FEEDBACK LOOPS: WHAT THE PRACTITIONERS SAY Feedback loops in behavioral economics, epidemiology and psychology etc.: The concept is used to refer to the re- sponse mechanisms by an individual, group or larger societal system based on four distinct steps of: (i) information input or ‘trigger’ (data, a specific experience, story, etc.), (ii) relevance and emotional connection to that information input, (iii) understanding of consequences with options for behavioral response, and (iv) action. It links use of informational triggers to a behavioral change or adaptation. Assessing and reflecting over such behavioral responses (at individual or system levels) can lead to repeat behavior if successful – either maintaining the adapted behavior or relapsing into the earlier patterns. The ‘information trigger’ at the beginning of every new feedback loop is different from the actual feed- back, which refers to the response mechanism itself and how that information was processed and responded to (by the individual, the collective, or the system). Feedback loops in social accountability and service delivery: ‘Feedback’ can be one-off or cyclical, and in many devel- opment interventions refers to the interactions between service provider and service user around needs or the satisfac- tion with a particular service or output. Simply put, ‘feedback’ mechanisms are about listening to the experiences and preferences of the people who are expected to benefit from change efforts. ‘Closing the feedback loop’ refers to ser- vices being responsive to user feedback with dialogue or engagement mechanisms in place for dialogue (taking it from one-way feedback to two-way communications, engagement, and delivery on mutual commitments). A more advanced interpretation of feedback loops includes the notion of shifting power dynamics between organizations/authorities and their primary constituents. For a good overview of different uses of the term, see this blog by Irene Guijt: http://bet- terevaluation.org/blog/feedback_loops_new_buzzword_old_practice Feedback loops in project or program monitoring: The notion of feedback loops is also central to monitoring and evaluation where learning-oriented monitoring and evaluation typically seek to build cycles of internal and stakeholder learning and reflection into ongoing monitoring efforts, seeking feedback from relevant stakeholders. External validation of feedback collected on a more ongoing basis in the monitoring is then done through mid- and end-term evaluations. Some may also refer to this as ‘learning loops’ (see main text), as it is closely related to the institutionalized mechanisms for organizational performance and learning. IV. AN EVOLVING FIELD- SOME TRENDS The focus and renewed interest in adaptive ed upon. This represents a more classical model management for development has shifted the of extracting and then (much later) disseminat- learning discourse from learning as a ‘good ing findings once they have been ‘packaged’ thing to do’ (but perceived by some as a bit of and analyzed by a smaller group of people, a luxury when dealing with pressing develop- often experts. Even in development research, ment needs) to linking it more clearly to per- what goes into such analyses are being increas- formance and results – i.e. learning as a driver ingly stored on open sources, using open data for change and impact at scale, rather than as formats or is more widely accessible so that an add-on or afterthought to operations. This there could be multiple interpretations using clearly has an impact on how we perceive and the same data set, with opportunities for ‘so- do knowledge management as well. cializing’ knowledge and information along the way. For one, the increased demand for real-time learning and experimentation means that we This trend of ‘democratizing’ data usage, blend- are increasingly moving away from the notion ing it with citizen feedback and perception polls that knowledge can be managed or stored on specific services, is an area where develop- centrally, then disseminated and magically act- ment agencies’ (including the World Bank’s) /7 open data, open governance processes and so- recently researched the World Bank’s own learn- cial accountability converge – particularly when ing practices, reflecting critically about how to it comes to filling gaps for granular data at sub- become a more adaptive learning institution national levels. There is scope to expand on this (see Section 6). The challenge will be to connect in the context of involving citizens in providing meaningfully this renewed investment in ‘front- feedback – not just on services as users – but line (or grassroots) learning’ with mechanisms on some of the underlying barriers to inclusive for adaptive management based on learning service access and use. strategies in larger development institutions. Another trend is that – contrary to the past – As a sector, however, our knowledge manage- many funders (philanthropic and international ment approaches and tools still are not at par aid agencies) in the social accountability field with the complexity of problems we are trying in particular are truly interested in investing in to fix. Therefore, we need to continue to better learning as part of the implementation process. understand the full spectrum of knowledge ap- There is also overall recognition of the fact that proaches and functions, as well as how they re- such funding streams need to reach those at late to different aspects of social accountability. the frontline of action, where learning-by-doing This could schematically be described as span- is happening close to the ground and where ning from simply managing information (build- ‘collective voice’ through organized groups of ing up your own or your institution’s knowledge civil society is in the focus. This is reflected both repository), to disseminating information, so- in the new strategies of the Ford Foundation cializing and interpreting information through and the Hewlett Foundation where it is linked exchange, through to experimenting with how also to their own learning agenda (see quote it is possible to relate to and finally adapt be- box). haviors and attitudes in concrete action (see Ta- ble 1 below). ”We want to understand how best to sup- port subnational groups, such as teachers’ V. A WIDE SPECTRUM OF APPROACHES AND LEARNING LOOPS and parents’ associations, youth groups, women’s organizations and school man- Albert Einstein has allegedly and famously said agement committees; how to avoid having that imagination is more important than knowl- such groups captured by elites; and how edge since “knowledge is limited to all we now these groups are (or are not) engaging in know and understand, while imagination em- useful ways with national-level civil soci- braces (…) all there ever will be to know and un- ety organizations.” derstand.” A parallel can be drawn to the area Hewlett Foundation, Transparency, Participation & Ac- of knowledge and learning – knowledge being countability Grantmaking Strategy, Dec. 2015 about what we know, or know that we don’t know (with training to fill those gaps). Learning From a donor perspective, the need for itera- on the other hand, involves a more complex set tive, context-relevant learning is also reflected in factors to unlock what we do not even know the Smart Rules for Better Programme Delivery11 that we don’t know – much like imagination adopted by the United Kingdom’s Department is needed to push the limits of what is known. for International Development. The Independ- Acting on this knowledge takes us even further ent Evaluation Group of the World Bank has also into behavior theory. 11 DFID, UK, 2014. See: https://www.gov.uk/government/publications/dfid-smart-rules-better-programme-delivery /8 Yet, in many organizational contexts, these two concepts are usually lumped together managerially without trying to understand where across the range of approaches we operate, and for what pur- pose. A somewhat simplified classification is outlined in the table below. WHAT Knowledge Knowledge Knowledge ex- Learn to relate Learning to adapt hoarding (for dis- generation (using change (socializa- (preparation) ( experimentation) semination) existing data) tion) HOW Amassing infor- Interpretation and Making ‘Aha Relating informa- Trial and error mation and mak- sense -making moments’ pos- tion and knowl- ing it availlable sible leading to edge to own reinforcement concept or unexpected revelation MAIN TOOLS Data-bases, Research Networking, Adult education Behavioral portals, gate- interpersonal principles, case- feedback loops ways, knowledge and peer-to-peer based discus- (evidence, action, repositories exchange, focus sions, experience- consequence, group validation based reaction) FOR WHOM? Unknown Cluster of inter- Usual or ‘unu- Groups Individuals and/ ested audience sual’ suspects or interconnected interacting, often (complex adap- facilitated tive) systems MOTTO ‘It’s all about the ‘It’s all about the ‘It’s all about ‘It’s all about the ‘It’s all about data’ analysis of data’ the interaction relevance of data input, action, around the data’ put into practice assessment & and context’ reaction’ Table 1. ‘Simplified’ overview of a range of knowledge & learning approaches Looking across the various approaches out- The recent focus on adaptive management and lined in Table 1, it is clear that we need all of learning also cuts across the different categories these categories, sometimes simultaneously, outlined above since knowledge generation or sometimes in a phased way, or adapted to some form of information trigger is usually nec- different settings and learning needs among essary for engagement between actors to take our various stakeholder groups. For instance, off in joint processes of problem-solving. ‘knowledge hoarding’ for later dissemination without engaging in some socialization of the Even so, many knowledge management initia- information could be seen to drastically reduce tives tend to get stuck primarily in one category the usefulness of the information to local stake- or another (in line with the ‘mottos’) or look at the holders and peers. Likewise, trial and error on various actors involved in social accountability its own without combining it with a feedback processes and divide up the tasks rather artificially. and an evidence-based tracking approach One potential trap is that ‘upstream’ funders could be seen as wasteful and irresponsible. take on the task of amassing large datasets /9 that are divorced from language and context ful to analyze how different information stimuli that is relevant for implementation, or commis- may or may not lead to an alternative behav- sion less timely research pieces, while we look ioral response, or adaptation. A feedback loop to frontline practitioners to do the learning-ori- is simply a loop (small or large) of action, in- ented adaptation in a vacuum of useful knowl- formation, and reaction. When the reaction is edge support. measured or assessed, a new loop begins and so on. Positive feedback loops lead to a rein- Given that we operate in aid management forcement and alignment of interests towards a frameworks where learning is ‘projectized’ rath- constructive common goal or agenda (even if er than seen as a continuum towards longer actors’ motivations or incentives to achieve that term change processes, important ‘learning goal differ), whereas negative feedback loops journeys’ are rarely invested in, nor document- can lead to increasing divisions between actors ed. Moreover, traditional knowledge manage- or sub-groups, tensions and ultimately chaos ment tends to confound learning with a series and backlash (taking a certain change agenda of knowledge products, where information is backwards rather than forwards towards posi- ‘managed’ (in large knowledge repositories, tive results achievement). Also, all feedback databases or ‘interpreted’ by experts before loops can be said to comprise four stages: in- it is released) rather than being put to instant formational input or evidence of some sort (not use for continuous learning at multiple levels of necessarily numerical), relevance, consequence engagement. Learning-oriented adaptation, on and action in systems of human interaction12. the other hand, puts a far bigger emphasis on Similarly, the idea of ‘learning loops’ applies the information usage and action. ideas of iterative ‘loops’ to individual or organi- zational (or ‘system’) learning and are geared to An adaptive approach to learning furthermore achieving specific results. acknowledges that lots of small and incremen- tal changes at multiple levels, by multiple ac- These ideas are also not new. In fact, the con- tors, can accumulate into a tipping point that cept of ’single’ and ’double loop’ learning was challenges existing ‘holding patterns’ (such as developed by Chris Argyris already back in the power, politics, gender roles etc.) that blocks 1970s, and have later been added to and fur- change in a system. Consequently, learning and ther developed in different contexts. In adaptive adaptation has to happen at multiple levels si- management and learning, it’s worth revisiting multaneously through a series of interwoven re- these concepts to better understand how differ- lationships, and inspire deeper transformational ent types of learning can occur, and to be able processes of learning that can also question to adapt the knowledge management response the current way things are done, the operating accordingly. Rather than just talking about ‘clos- framework and ‘rules’, as well as the dominant ing the feedback loop’ (though a response or mental models and discourses (see reference to action), it connects this to a transformational single-, double- and triple-loop learning below). learning process in order to achieve results. Single, double and triple-loop learning As mentioned, ‘feedback loops’ – the way it is being used in behavioral disciplines – is use- 12 There is an extensive literature on this related to behavior theory. Another resource is ‘Triggers’ by Marshall Goldsmith and Mark Reiter, Crown Publishing Group, New York, 2015 / 10 BOX 2 “’Single-loop learning’ refers to learning and corrective action within the same goal-structure and rule-boundaries. This is a simple feedback loop, where outcomes cause adjustment of behaviors, like a thermostat. It is generally in operation when goals, beliefs, values, conceptual frameworks, and strategies are taken for granted without critical reflection. A higher order of learning is when the individual questions the goal-structures and rules upon detecting an error. This is more like ‘coloring outside the lines’ to solve the problem or error. This is referred to as ‘double loop learning’. This is more creative and may lead to alterations in the rules, plans, strategies, or consequences initially related to the problem at hand. Double-loop learning involves critical reflection upon goals, beliefs, values, conceptual frame- works, and strategies.” Source: www.lifecircles-inc.com/Learningtheories/constructivism/argyris.html A third level has also been added to widen the also illustrated in the graph below, could be: single, and double loop to a ‘triple loop’ where learning is embedded in context and ques- Single loop learning: ‘Are we doing things right?’ tions how current rules and priorities are being (within the boundaries of our current frame- set. This is a larger, system-wide feedback loop work) where existing holding patterns and power Double loop learning: ‘Are we doing the right inequalities come into play, and where exter- things?’ (or do we need to change processes or nal stakeholder feedback can help as a reality strategies?) check to stay grounded. There are many ways Triple loop learning: ‘How do we decide what is to interpret this approach and to apply it to or- right?’ (in the unique context in which we op- ganizational, project-specific or ‘system level’ erate, what are our current basis for decision- learning. Some key questions for each stage, making?) Context Assumptions Actions Results Triple loop learning= Double loop Single loop How do we decide learning= learning= what is right? Are we doing the Are we doing right things? things right? Figure 1. Single, double and triple loop learning linked to results / 11 The above figure seeks to illustrate that working Knowledge and learning for adaptive manage- in adaptive systems involves learning at all levels ment needs to be closely aligned to a project of implementation in an iterative process, and is or organization’s own performance monitoring, closely linked to managing for results. This goes embedding learning loops close to the ground beyond just project-specific corrective learning for quick stakeholder feedback in order to re- (are we doing things right?) or programmatic move uncertainties and expose ‘blind spots’ learning where we question underlying as- (what we don’t know that we don’t know) in re- sumptions or strategies (are we doing the right lation to achieving the desired results. Learning- things?). Regularly scrutinizing how we decide oriented monitoring seeks to focus on learning what is right in a given context, means tapping at multiple levels by asking the question ‘who into different streams of knowledge, challeng- needs to change (and learn) how in order for ing existing mental models and listening to desired change to happen’ (including the im- feedback from the ground as well as partners plementers themselves). In learning-oriented and non-partners (the use of ‘unusual suspects’ approaches – such as for example Outcome have been used as a break-away from always Mapping and Outcome Harvesting – attitude talking to the same people, instead listening to or behavioral change, rather than activities and those who are likely to have a different opin- outputs, are the focus of any internal tracking ion or who are unexposed to dominant ideas mechanisms. By merging learning loop think- and thought frameworks). Clearly, this needs ing with the role of knowledge management, to be closely linked to existing monitoring and we can better track how to use of information, evaluation processes to link learning and adap- knowledge and learning in order to increase tive management with results frameworks and the uptake and external response to a specific monitoring practices. intervention. Looking at existing practices of monitoring and evaluation (M&E) against this learning para- Making monitoring the ‘learning engine’ of digm, it has been argued that while there will any externally funded initiative should be a priority… always be a dimension of upstream account- ability to donors (are we doing things right?) and a dimension of downstream accountability to end-users (are we doing the right things?) Without such flexible learning mechanisms based on assumptions that needs to be tested, in place that will systematically feed informa- a third dimension is that development agencies tion into adaptive management processes, you increasingly are asked to be accountable also to could say that whole project cycle becomes one partners and peers in what has been referred big behavioral feedback loop with little scope to as ‘triple accountability’13 . The fact that de- or room for adaptation before it is over (except velopment agencies are increasingly opening for possibly in relation to a mid-term evalua- up its own data processes and incorporating, tion, but by then there is already a lot to lose if systematically, stakeholder and citizen engage- it comes out negatively). Re-thinking monitor- ment and feedback mechanisms about its own ing to become the ‘learning engine’ of any ex- performance can be seen as part of this trend ternally funded initiative should instead be the of ‘shared accountabilities’ across partners – priority. national and international. Engel, P 13 ., Keijzer, N., Ørnemark, C. 2007. ‘Responding to change: Learning to adapt in development cooperation’. (Policy Manage- ment Brief no. 19). Maastricht: ECDPM. http://ecdpm.org/publications/responding-change-learning-adapt-development-cooperation/ / 12 sus can be an obstacle to innovation and para- VI. LEARNING TO ADAPT WITHIN lyze critical self-reflection and learning. THE WORLD BANK There is a pronounced commitment to knowl- Nevertheless, many of the staff who partici- edge generation and exchange by the World pated in the IEG internal survey believed that Bank Group’s (WBG) leadership, as demonstrat- the World Bank’s role in facilitating contacts as ed in the desire to “capture all the best experi- a knowledge broker between countries was ences from around the world and then putting working well, particularly in some regions. On that information in a form countries can use the topic of using knowledge generated out- and try in their own local settings to improve on side the Bank for learning and for providing their own service delivery.” This was explained ‘cross-sector knowledge’ to meet client needs, by President Kim in a 2013 interview (UN News the picture was less clear, however. And while Centre 2013), reflecting his conviction that one most interviewed staff believed that it should of the central responsibilities of the World Bank be possible to adapt based on lessons within as a development agency is to increase the flow the span of a single project, there was overall of knowledge and learning across the world. agreement that intra-project adaptiveness was Work on learning from ‘trial and error’ in lend- lacking. Some also noted that most Bank-gen- ing operations that are more adaptive have fed erated knowledge was focusing on the ‘what’, into this line of thinking14, along with efforts to as in ‘what works’. Yet “the solutions that gov- more effectively ‘harvest’ learning from inter- ernments seek are often operational in nature nal communities of practice that span across – they are about the ‘how’, not the ‘what’”.16 and go beyond individual projects. Implement- ing this vision in practice may be more difficult, “The only way to scale impact is to make however. regular people do extraordinary things. For that you need to develop the systems that allow them to learn.” In 2015 the Independent Evaluation Group of the WBG published a substantial research piece15 NY Times columnist Tim Friedman at the launch event of aimed at better understanding internal learn- the World Bank Open Learning Campus, 12 Jan. 2016 ing and adaptiveness in its lending operations. It found, among other things, that Bank staff ‘Safe space’ for questioning and critically learn are keenly aware of the importance of infor- also seemed to be largely lacking since less mal learning and group work, often referring than one third of the IEG survey of Bank staff to mentoring, learning from peers and using felt that they could openly discuss with their personal networks. While this could be a good management what is not working in a lending thing, it also tended to reinforce mindset biases, operation. In one case where successful adap- with a perceived difficulty of challenging status tation was noted, it was led by strong Bank fa- quo since informal learning involves a large el- cilitation and responsive high-capacity partners ement of copying others’ behavior. This is also combined with innovative use of mechanisms where external feedback and other sources to harvest ideas from field staff. In another case, of evidence can help to off-set a paradigm of the project departed from each of the good ‘group think’—i.e. a state where group consen- practice principles which World bank studies 14 Andrews, Pritchett and Woolcock, 2012. 15 Learning and Results in World Bank Operations: Toward a New Learning Strategy, IEG, 2015. 16 IEG, 2015, p.63. / 13 and guidance had recommended as a blue- ously test their own assumptions against actual print, instead letting the local context drive the learning of key stakeholders. Rather, proposals 18 need for flexibility.17 The IEG report lead author generally contained activities geared at how also stressed that Bank staff perceived the lack others could learn from them. of institutional incentives as one of the biggest problems for learning and knowledge sharing, Even though the GPSA grant application for- calling for a system of both financial and non-fi- mat specifically asked applicants to develop a nancial rewards to ensure learning behaviors at rationale for knowledge and learning that is scale. This, she meant, should go hand in hand linked to their own learning, integrated in on- with sufficient time and budgets set aside for going program management practices (such as learning and knowledge sharing. in their monitoring and evaluation), there was a tendency to reduce knowledge and learning And in the context of the GPSA to a set of training tools, workshops or publica- tions of ‘best practice’ (rather than the learning The Global Partnership for Social Accountabil- process of determining what makes such prac- ity (GPSA) was designed to invest in and link tice the ‘best’). This was, however, not surpris- practitioner-based learning from its civil society ing since donors are often at fault for reducing grantees19 and partners to broader processes knowledge and learning to a set of pre-deter- of sector-wide and institutional learning across mined activities and tools rather than a focus the World Bank20. Capacity-building is pro- on learning outcomes for different audiences, vided to the civil society grantees to continu- including implementers and organizations ously share knowledge, self-assess and adapt. themselves. Showing ‘what works’ and how To date, the concept of ‘adaptive learning’ has they are certain about it, seemed to be instead primarily been used to refer to grantees’ abil- what many of the civil society organizations ity to continuously course correct in their indi- from around the world were used to include in vidual projects, linking learning closely to results proposals submissions.22 achievement. This was explained in an earlier GPSA Note21 where a sample of 40 of 644 ap- The grantees from the first couple of rounds of plications were analyzed to see how applicants Calls for Proposals (2012 and 2013) were, sub- responded to the question of how they would sequent to their grants having been approved, incorporate learning for improved results. In line asked to draw up individual Knowledge & with subsequent proposal assessments, it illus- Learning (K&L) plans including different aspects trated that many grantees faced challenges to of knowledge management practices such as fully grasp the link between learning and results knowledge generation, knowledge sharing, ex- and to turn this into actionable components for change visits and feeding learning continuously the knowledge and learning provisions in the into internal planning processes and reviews as budget. Overall, few proposals reflected learn- well as into GPSA activities (knowledge platform). ing-oriented monitoring practices to continu 17 The cases refer to WB loan to the Department of Social Welfare in the Philippines and to small enterprise access to finance project in Turkey. IEG, 2015, p. 65-66. 18 Soniya Carvalho. 19 Currently funding 23 grantees in 17 countries based on two Calls for Proposals (2013 and 2014), with a third Call completed in early 2016. 20 By, among other things working closely with the Citizen Engagement Secretariat for Bank operations. 21 Gurezovich, F., Poli, M. : Are We Ready for Strategic Social Accountability? Note 5: Adaptive Learning 22 Ibid. / 14 An internal stock-taking exercise of the K&L government counterparts at GPSA grantee work- plans in 2015, however, found that although shop in 2015, who they thought would learn most quality of these plans varied, overall there from their GPSA projects, most agreed that it were gaps in the area of making projects more would be the project implementation team itself learning-oriented by integrating various learn- (40%). Less than one in five thought beneficiaries ing and knowledge approaches into the work. would be the main learners with ‘government’ be- Rather, learning was often seen as a separate ing seen as the least likely to learn the most from ‘add on’ to the main operations. At the same the experience. While this is hardly surprising, it time, much of the practical learning took place opened up for an interesting discussion about how through day-to-day implantation When asked to broaden the spectrum of learning-by-doing in during an interactive polling of around 70 rep- actions that are dependent on mutual learning resentatives from grantees and some of their and co-creation of solutions among several actors. Who do you think will learn the most in the course of the project? 40% 17% 14% 9% 11% 3% 6% 1 2 3 4 5 6 7 1. Project implementation team 5. Government 2. Knowledge and learning staff 6. GPSA team 3. Beneficiaries 7. World Bank 4. Project Partner CSOs Souce: Poll taken at GPSA Grantees Workshop, May 2015 / 15 A bigger lesson was also that a specific budget allocation for knowledge and learning – which VII. SOME PRACTICAL STEPS FORWARD is a good start – is not enough for shifting to- wards more adaptive management practices An often stated pre-requisite for ‘group’ and based on mutual learning alongside others organizational learning is that questioning is without providing clear guidance as to how safe, that learning is welcomed by senior man- GPSA frames the concepts of knowledge and agement, and that learning-oriented behavior learning for improved results, with an emphasis is rewarded. That would be ideal. The reality on learning alongside others for more adaptive in many (if not most) settings, however, is that management, with other knowledge products conditions for learning are not perfect. ‘Think- to support the behavior changes they seek to ing outside the box’ among employees may obtain (e.g. a more responsive service provider hypothetically be welcomed, but inadvertently to citizen’s needs). A shift to focus more on how discouraged in practice. And yet, amazingly, to document and support grantees’ ‘learning adaptation, evolution and change happens all journeys’ towards adaptive management will be the time in large as well as small organizations. explored in response. Given that learning is a continuous process that happens at multiple levels, what might start out Importantly, the learning potential of the World as individual learning may attract a cluster of Bank in the field of citizen engagement and like-minded. Improvements in performance or social accountability has also come into focus results may win other stakeholders over. since the WBG’s Strategic Framework for Citi- zen Engagement was adopted in in FY 2013/14, Innovative knowledge management can making citizen engagement mandatory in all help connect otherwise parallel information World Bank financed operations where benefi- or evaluative streams… ciaries can be clearly identified. This has opened up a new internal demand for learning in this There are several concrete ways in which grant- field, which the World Bank’s Citizen Engage- ees and donors alike can seek to make pro- ment Secretariat and a network of citizen en- cesses more learning-oriented and adaptive. gagement focal points across Bank operations An important place to start is to add an explicit are trying to address, and with whom the GPSA learning dimension to operations, including the is collaborating as an internal knowledge and theory of change, along with existing monitor- learning resource. ing practices and frameworks, but also to in- creasingly use innovative knowledge manage- Increasingly ‘connecting the dots’ between the ment practices to connect otherwise parallel GPSA, the citizen engagement agenda and information or evaluative streams. Some ideas open governance is being used as a strategy for possible action areas are outlined below. to maximize learning – from those at the front- line to the sector as a whole. However, more can still be done, particularly around how continuous learning and adaptation for op- timal social accountability outcomes can be achieved, and how this in turn can help improve governance effectiveness in specific sectors. / 16 In relation to intervention design& set-up Also, if our aim is to influence the development sector more broadly, we also need to map out how we intend to serve existing ‘learn- Make learning assumptions explicit ing needs’ at that broader, international level Interactively analyze a particular project or (which is different from assuming that ‘good longer-term initiative’s intervention logic and practices’ will spread and scale on its own). overall change agenda to reveal and spell out learning objectives and assumptions around learning (for whom, how, with what effect) for Regularly adapt plans based on changing each deliverable or milestone. To do so, iden- and emerging learning needs tify (i) the learning assumption that is associ- ated with each expected behavioral change In doing so, it is important to keep in mind that among different stakeholders, (ii) whether the actual learning needs (e.g. what internal or ex- management systems are suitable for regularly ternal actors will need to know or learn) is not monitoring and nurturing such learning, and necessarily known at the outset of a project or (iii) whether outputs are likely to be context- process. This also means that knowledge and relevant and significant enough to trigger a learning plans need to be flexible enough to positive adaptive response by relevant stake- accommodate for changing needs and put a holders (based on evidence, action, assessment bigger emphasis on information uptake and us- and reaction), and (iv) how this aligns (or not) age at different levels. Often, this may mean with the overall change theory of the project. a shift away from more predictable or ‘tradi- tional’ knowledge outputs such as publica- By mapping out implicit learning assumptions tions or data-bases unless it is clear how they related to key outputs (by individuals, organiza- will be used. Likewise, capacity needs assess- tions and the overall stakeholder system) it is pos- ments, useful as they are, tend to focus on sible to track and verify such assumptions along known, rather than unknown capacity gaps at the way. Otherwise it is possible to end up with a the outset of a process. Adapting in systems situation where all deliverables have been pro- of complex change means that learning needs duced, but where no real learning has occurred. will change based on the actions of others Mapping out the expected (and real) ‘learning in the system. In the logic of adaptive man- journeys’ in longer-term processes can also help agement, capacity needs assessments would to highlight over-simplifications of how learning therefore need to be regularly revisited and and behavior change occurs, and by whom. linked to monitoring of actual learning out- comes as well as of emerging learning needs. For instance, if we assume that ‘citizens’ will learn to hold public officials to account, it is In existing knowledge & management important to break down the different learn- practices ing and information needs among different sub-groups of citizens and to test if our in- Review knowledge management strategies tervention can provide them with relevant in- from a learning perspective formation triggers that can lead to increased knowledge and learning along the way. Con- A review of existing knowledge manage- versely, are the informational triggers and ment practices and outputs could inquire into learning process for targeted public officials whether they actually lead to learning and drive relevant enough for them, and are the conse- change forward – internally and/or externally. quences for e.g. not responding to citizen feed- For instance, is the primary knowledge ap- back clear enough to help cause a response? proach stuck in the mode of ‘amassing knowl- / 17 edge’ in large databases or producing publi- What can we do to make information trig- cations that will at some point be shared and gers more relevant? What about the timeli- hopefully inspire learning by others? Or is the ness and abilities for stakeholders to put their project using knowledge generation, shar- new knowledge into practice? How will we ing, socialization and internalization strategi- test and learn from what we think will work? cally to help achieve key outputs? (See Table It would also be important to closely align 1). Is the balance right between targeting ex- learning objectives and their monitoring ternal and internal stakeholders for learning? with existing results frameworks, in order to Another question could be whether learning avoid setting up parallel processes internally. happens at more than one ‘level’ of opera- tions, and how do we know? While ‘operational learning’ may be reflected in the regular re- porting, our influence and relevance to peers In relation to current monitoring and and the sector as a whole is often assumed. evaluation (M&E) practices Clearly articulate the knowledge manage- Link learning to monitoring & evaluation ment approach upfront and organizational strategy At organizational level – or in view of a particu- Having learning-oriented monitoring mecha- lar intervention or project – clearly articulating nisms in place would be a helpful first step to- the approach to knowledge management and wards integrating a learning dimension to what its role in relation to multi-stakeholder learning can otherwise be a routine practice of data (internal and external) can be helpful. This can hording in M&E databases. But we also need help to articulate how knowledge and learning an ‘inquisitive mindset’ to make monitoring and is integrated into operations, how it comple- learning part of day-to-day operations. If learn- ments and connects knowledge streams across ing is seen as an important part of the organi- stakeholders, and how it is being used to maxi- zational strategy and something that the senior mize knowledge uptake that leads to action at leadership puts emphasis on, it will help to cre- multiple levels in broader systems of change. ate this ‘culture of learning’ which is closely linked Clearly stating the organization’s ‘philoso- to performance improvements and results. phy’ about knowledge and learning may also help internal and external stakeholders to be One suggestion would therefore be to start more constructively self-critical and learning- with the strategic plans, operational plans and oriented in its interactions and/or reporting. work plans, and see how and where learning is part of the regular reporting and tracking in- ternally. Is it rewarded, is it mentioned, is it vis- Clearly spell out and monitor learning ob- ible? Are there openings and opportunities for jectives alongside each output and outcome feeding learning back into the organizational memory? Both monitoring and evaluation will Often, the learning objectives are unclear have multiple functions, but which function is or not spelled out, and monitoring of learn- dominant in a particular project, organiza- ing outcomes as a part of overall progress tion or setting: the accountability function, the and performance monitoring is often lack- learning function or the communications/fun- ing. Monitoring of learning objectives would draising function? How ‘narrowly’ are moni- need to go beyond assessing the extent to toring results shared and discussed internally? which external stakeholders are learning, As for external programmes and results frame and inform also our own role in that process. works – even if they are rigid in their format – what / 18 would be the ‘learning dimension’ of each in- It has argued that the ‘ability to act on informa- dicator to be measured? Who would gain tion’, as an important complement to informa- from that learning? How can things be tion access and the production of knowledge adapted to widen the learning circle to in- products that are largely extractive, should be a clude more stakeholders – e.g. by shifting growing concern for our knowledge manage- the learning paradigm closer to the ground? ment (KM) practices. It could be particularly rel- evant in the field of social accountability where progress is dependent on incremental atti- Separate Monitoring from Evaluation – and tude and behavioral changes of many actors. make ‘monitoring’ about learning… Monitoring and evaluation could be seen as Another key concern is to understand what type two sides of the same coin. However, often of information or knowledge input is needed stringent reporting formats or results-frame- when, for whom, to contribute to the desired works that correspond to donor’s account- changes outlined in our theories of change. A ability needs – rather than to the learning key question to ask would be ‘who is expected to needs of implementers or users – mean that act on this information how’ for each knowledge the M&E process usually is linked neither product or process, knowing that timeliness to learning, nor to adaptive management. and relevance is needed to close any behavio- One way around that could be to set up an ral feedback loops. Some feedback and behav- organizational monitoring system that re- ioral change loops are instant – often relating to sponds to both donor requirements and in- a simpler decision-making process with fewer ternal learning needs, using innovative ways actors involved or dependent on the outcome. of involving those closest to the ground (and Such incremental changes could be important thus most likely to adapt through learning-by- to track and build upon as they can accumulate doing) in regularly tracking key behavioral or to larger shifts in attitudes, discourse or behav- attitude changes where it most matters. Build- ior. Others take place at more systemic level be- ing and rewarding learning ‘from below’ at the fore leading to concrete change on the ground. frontline as part of the delivery mechanism, is something that funders may well both accept Internal monitoring and evaluation efforts and wish to invest in. So while monitoring and tend to track project outputs and keep track evaluation should stay linked, it may be help- of tangible knowledge outputs produced. To ful to look at monitoring as the internal learn- more effectively use and integrate knowl- ing aspect of what will periodically be validated edge management, however, such behav- (or contested) through external evaluation. ioral and attitude changes – if they were in- tegrated into learning-oriented monitoring – could help us better understand informa- tion uptake, learning, knowledge and adapta- VIII. CONCLUSION tion much more consistently across operations This note has looked at the role of adaptive man- The paper has suggested that the discussion agement in the context of how we use knowl- around behavioral change and feedback loops edge and learning in processes of complex is relevant in the context of adaptive manage- change (as opposed to interventions with high ment and learning as it illustrates the need to causal predictability), particularly focusing on constantly test assumptions around behavioral the field of social accountability, and drawing on response at different levels (individual, organi- some observations and lessons from the GPSA. zational, system). It is relevant in the field of so- / 19 cial accountability for a couple of reasons: (i) to see how we most effectively use ‘infor- mational triggers’ (such as citizen-driven data and other data sources) to put into motion incremental behavioral responses among all stakeholders required for horizontal and verti- cal social accountability systems (from citizen, government through to international donor agency), and (ii) to understand how knowledge management can better support the adaptive process among multiple stakeholders involved in citizen-driven social accountability. Finally, it has sought to take the discussion around learning away from the more insular field of in- trospection to one where it is framed as a key driver of change and managing for results, with some suggestions for practical ways forward. / 20