Are adaptations self-organized , autonomous , and harmonious ? Assessing the social – ecological resilience literature

The paper analyzes how adaptability (adaptive capacity and adaptations) is constructed in the literature on resilience of social–ecological systems (SES). According to some critics, this literature views adaptability as the capacity of SES to self-organize in an autonomous harmonious consensus-building process, ignoring strategies, conflicting goals, and power issues. We assessed 183 papers, coding two dimensions of adaptability: autonomous vs. intentional and descriptive vs. normative. We found a plurality of framings, where 51% of the papers perceived adaptability as autonomous, but one-third constructed adaptability as intentional processes driven by stakeholders; where social learning and networking are often used as strategies for changing power structures and achieving sustainability transformations. For the other dimension, adaptability was used normatively in 59% of the assessed papers, but onethird used descriptive framings. We found no evidence that the SES literature in general assumes a priori that adaptations are harmonious consensus-building processes. It is, rather, conflicts that are assumed, not spelled out, and assertions of “desirable” that are often not clarified by reference to policy documents or explicit normative frameworks. We discuss alternative definitions of adaptability and transformability to clarify or avoid the notion of desirability. Complex adaptive systems framing often precludes analysis of agency, but lately self-organization and emergence have been used to study actors with intentions, strategies, and conflicting interests. Transformations and power structures are increasingly being addressed in the SES literature. We conclude that ontological clashes between social science and SES research have resulted in multiple constructive pathways.


INTRODUCTION
Research on resilience of social-ecological systems (SES) originates from ecology and complex adaptive systems (CAS), focusing on ecological resilience, tipping points, feedbacks, selforganization, and other system properties (Berkes and Folke 1998).The recent shift in focus from resilience of ecosystems to resilience of SES has attracted-and has partly been driven bysocial scientists, especially on adaptation and transformation issues.In social science, both adaptation (as action) and adaptability (adaptive capacity) concern agency, i.e., real actors with intentions and conflicting strategies in specific contexts (Nelson et al. 2007).Social science frameworks, therefore, often clash with CAS frameworks, where adaptations in ecosystems are seen as emergent properties of local interacting but autonomous parts (Levin 1998).
There has been a heated debate on whether the research field of SES resilience is still based on ecological ontologies or whether it has changed into a truly interdisciplinary research field that enhances understanding of the coevolution of intertwined social and ecological systems (Folke et al. 2010).According to some authors, this change has already occurred through significant advances in analyzing agency, especially in relation to transformations, focusing on learning, innovation, leadership, and "changes to practices, lifestyles, power relations, norms and values" (Brown 2014:113).Power, agency, and justice issues belong to the priorities for future SES research (Olsson et al. 2014, Fischer et al. 2015).Other authors emphasize the lack of change (Leach 2008, Hornborg 2009, Hatt 2012), claiming that the SES literature views resilience and the necessary adaptations and adaptability as autonomous (spontaneous) and harmonious, assuming consensus building.Nelson et al. (2007:409) argue that the idealized situation-of testing and revising management and institutions based on ecological knowledge in a self-organized process-is rare, and that this approach may reinforce existing inequalities if power relationships are ignored.Cote and Nightingale (2012) suggest that the shortcomings in addressing power relationships and conflicting interests in SES research is due to an assumption that social and ecological systems are essentially similar.
Much of the critique of the SES literature can be codified into two propositions: (i) it uses a framing of autonomous selforganization, which precludes analysis of intentional agents, conflicts, and power; and (ii) it uses a normative framing.By "normative," we mean statements that assume an underling norm of what is desirable.Facts become normative when they "are interpreted through the filter of an assumption that implies an inherent policy preference" (Lackey 2001:439).
In the context of sustainable adaptations, Brown (2011:29) uses resilience as a normative concept, the opposite of vulnerability, and points to "the need for fundamental institutional reconfiguration in support of long-term equity and resilience."Hence, social change is necessary for resilience.Other authors make the opposite normative association, equating resilience with social stability (MacKinnon and Derickson 2013), and by associating it to functionalism, arguing "that resilience appears conservative when extended to social change and social relations" (Olsson et al. 2015:6).Assumptions of desirability in the very definitions of adaptability and transformability have also been criticized (Leach 2008).
However, the self-organization of CAS may not necessarily be employed to legitimize the present or any other desired social order; it seems rather strange to associate a scientific approach (resilience or CAS) to a political ideology.In a contrasting example, Brown (2014:113) discusses transition towns and argues that " [c]ounter to the argument about resilience as supporting regressive and neoliberal agendas, resilience is being used as an organizing principle by communities to challenge the status quo and to design and shape alternative futures." Central to this debate on self-organization are the concepts of scale and emergence.In CAS frameworks, actors are often observed from the outside, e.g., from a higher level of a spatial or institutional panarchy (Holling 2001).Such analytical frameworks tend to view the unfolding of events and adaptations made in the system as autonomous and emergent, with adaptation being an inherent self-organized property of CAS (Levin 1999).However, the interaction between agents that results in emerging properties need not be autonomous but may include "players and strategies" (Lansing 2003:196).In SES, self-organization is often used for processes occurring at lower organizational levels, for example, when Ostrom et al. (1999:281) observe that "[n]ational governments can help or hinder local self-organization" or when Folke (2006:260) suggests that self-organization is the opposite of either "lack of organization, or organization forced by external factors."If an ideological interpretation must be made, the two http://www.ecologyandsociety.org/vol22/iss1/art12/citations above may reflect a preference for decentralization and voluntary action rather than a preference for Hayek's spontaneous order.Indeed, self-organization in the field of SES research often refers to voluntary organization without assuming harmonious or autonomous processes.For example, Österblom and Folke (2013:2) employ a CAS perspective to study how global governance to combat illegal fishing emerged from interaction among key persons who managed to mobilize networks and formalize cooperation.Holling's (2001:403) argument that "Self-organization of human institutional patterns establishes the arena for future sustainable opportunity" makes sense from a larger-scale systems perspective, but clearly abstracts from the strategies, leadership, conflicts, alliances, and power imbalances of human actors.However, abstracting from these issues is not the same thing as assuming that adaptations occur harmoniously through consensus; indeed, it says nothing about how adaptations actually occur and gives no information on the strategies, power structures, conflicts, and leadership underlying adaptability.
If we want to understand "the art" of adaptations and transformations, i.e., the personal skills and networks of key change agents navigating the unexpected (Westley 2002, Westley et al. 2013), and not just the need for it, we need to analyze agency in specific contexts.As long as there are plural opinions of what ends society should promote, what means should be employed, and how trade-offs should be dealt with, there will be conflicting interests.This applies even in the case of global consensus on the ends, which the United Nations (2015) suggests exists for the new Sustainable Development Goals.There are often strong economic incentives for actors in SES to strategically plan, advocate, or simply coerce a different state or development trajectory than the one envisioned by the norm of success applied or assumed in the scientific analysis.For example, in a study of water governance in Sweden, Galaz (2005) showed that some actors blocked attempts for social learning and adaptations desired by other stakeholders, an act of obstruction of change linked to the institutional setup where hydropower concessions have been unlimited in time in Sweden (Rudberg et al. 2015).
Ideally, normative statements in science, e.g., assertions of what is desirable, should be evidenced by an explicit reference to a policy document or scientific-political framework with clearly expressed norms like the MA (2005b) or the new Sustainable Development Goals.For instance, claims that a clear lake is more "desirable" than a turbid lake are typically normative statements, but beg the question "for whom" or according to what policy framework?If clear lakes are desirable by all or most people in a society and lakes still become turbid, is it due to lack of knowledge or some actors driven by strong monetary incentives or something else?
In the context of natural resource management, an ecosystem that is approaching an "undesirable" threshold according to some actors and that is therefore not assessed as resilient in an ecological sense, may nevertheless be resilient in a social-ecological sense.This is the case if these actors show high adaptability and are able "to respond to change and restore the lake" (Folke et al. 2005:444).Hence, when shifting the perspective from ecosystems to SES, a more explicit reference to agency and normative issues concerning adaptability can be expected.
Functionalism was explicitly referred to only once (in Westley et al. (2002)) but not in the way the critics suggest; there is no preference expressed for maintaining or justifying the existing social order.Other texts (e.g., Berkes et al. 2003:364) discuss different "functions" or roles by key persons to accomplish a transformation: some act as visionaries, others as knowledge carriers, networkers, facilitators, entrepreneurs ,and implementers; again, not to maintain the existing social order, but rather the opposite.
We found no assumptions on consensus, but some texts described shared visions: "Trust-building dialogues [...] collaborative learning, and creating public awareness were part of the process.A comprehensive framework with a shared vision and goals that presented conservation as development and turned problems into opportunities was developed and contributed to a shift in values and meaning of the wetland landscape among key actors" (Folke et al. 2005:457).
The citation above by Folke et al. (2005) and the empirical papers it referred to suggest that consensus or harmonious processes are not assumed a priori.The processes for building trust, solving conflicts, and identifying common interests are studied empirically in the case studies described and involve key persons building alliances (networks) among diverse stakeholder groups.Although rarely made explicit, such networking is generally a strategy for changing power structures, and these are not smooth, harmonious processes (Hahn 2011, Crona andParker 2012).Olsson et al. (2014:4) admit that power is understudied in the SES literature but note that this is changing and that "the redistribution and sharing of power is one of the key conditions for more flexible, collaborative forms of management and governance that contribute to long-term resilience of socialecological systems."Thus, the SES literature cannot be said to assume harmonious consensus building.Conflicting interests are often implicitly acknowledged (networks are developed to change business-as-usual management and achieve a transformation), so it is rather the "conflicts" that are often assumed, not spelled out.
A high degree of consensus may be achieved in situations of "low politics" (Klijn and Skelcher 2007:596), i.e., when stakes are not high.However, voluntary collaboration is not a panacea, and the SES literature does not generally assume that a voluntary approach to collaborative adaptive governance is sufficient to achieve adaptations or transformation when stakes are high.As discussed before, the SES literature often acknowledges that major actors may obstruct any such process unless they are coerced or otherwise maneuvered to change.For example, turning one-third of the Great Barrier Reef into a marine protected area http://www.ecologyandsociety.org/vol22/iss1/art12/started as a collaborative learning process but eventually required political coercion to handle the opposition from fishermen (Olsson et al. 2008).Political coercion was also decisive for curbing illegal and unreported overfishing in the Southern Ocean but, interestingly, the coercion was enabled by a long process of trust building and collaboration between NGOs and civil servants in international networks (Österblom and Sumaila 2011).Thus, trust building, knowledge generation, and collaboration in networks can be an effective strategy for catalyzing government coercion and changing power structures.

METHODS
For the first two research questions, we used a structured literature review, an approach common to both social science (Petticrew and Roberts 2006) and natural science (Fink 2005).This part of the analysis was conducted in 2011-2012 (Nykvist 2012), a time when the strongest critiques of SES resilience were being published.We started by developing coding criteria for different conceptualizations of adaptability (Tables 1 and 2).We used the ISI Web of Science database and searched for articles published before 1 January 2011, matching ("social-ecological system*" OR "socio-ecological system*" AND resilience AND adapt*) in title, abstract, and keywords.We used the truncated expression of "adapt*" because we view adaptability as an inclusive concept including adaptive capacity, adaptiveness, and adaptations.The search yielded 193 papers, dating back to 2001.For the third research question, we made a simple quantitative assessment of all 798 papers on resilience, adaptability, and SES published 2001-2015, adding either the words "power" and "transform*" to the search.For each of these two added concepts, we did one analysis for every year to assess possible trends.The ways actors organize in order to adapt a SES toward a certain goal are described or analyzed.Actors can be assumed to have intentions or preferences in modeling.
The strategies or agendas of real actors are analyzed.
For research questions 1 and 2, each paper was searched for each occurrence of paragraphs containing "adapt*."These paragraphs were read, and key statements including definitions, descriptions, or discussions of adaptability were marked.Ten papers were excluded from the study because they actually had no discussions of adaptability in the text, despite passing the search criteria.We critically appraised the remaining 183 papers, analyzing the paragraphs including the word "adapt*" using our coding scheme (Tables 1 and 2) and we categorized the papers according to the two coding dimensions (Table 3).For research question 1, we were not able to code four further papers due to lack of definitions or descriptions.These papers are also excluded in our quantitative results.By having clear predefined categories and by discussing uncertain papers, we tried to minimize the inevitable subjective element of interpretation.Examples of our interpretations are provided in the results section.For more examples of coding and categorization of the 183 papers, see the Appendix (Tables A1-A3).It is important to understand that this method allowed no evaluation of the papers as a whole, only how adaptability was conceptualized and used.1).Some papers used both intentional and autonomous framings, and we coded these as "Both perspectives." Research question 2 concerns normative vs. descriptive framings.In a descriptive/analytical framing, adaptability refers to the capacity to adapt, whatever end or outcome.In explicit normative contexts, adaptability is the capacity to enhance the resilience of what is seen as a desirable ecosystem state or trajectory.However, sometimes there is no explicit reference to what is desirable, e.g., when adaptability is just assumed to be good or when adaptability is used in an instrumental way, expected to enhance resilience or learning without giving resilience or learning an explicit normative meaning.We refer to such statements as "vaguely normative" (see Table 2).Some papers used both normative (or vaguely normative) and descriptive framings.We coded these as "Several perspectives."Analysis of how actors organize and adapt with the aim to enhance resilience of a clearly defined desirable SES, or adapt and change the SES along a more desirable trajectory.
Adaptation is described as something positive, but reasons for this are not discussed.Actors and their intentions and strategies for adaptation are discussed.

Coding for the structured literature review
The 183 papers matching our search criteria were published in 62 journals (Fig. 1), of which Ecology and Society accounted for onethird of the papers (56/183).

Autonomous vs. intentional
More than half of the papers (51%) explicitly or implicitly refer to adaptations as autonomous only (Fig. 2A), often in the context of a complex adaptive system (CAS), or explained through the four phases of the adaptive cycle (Holling 2001).This set of papers thus provides no information on intentions, strategic planning, leadership, or power in relation to adaptability.For 31% of the papers, intentions and strategies are discussed in relation to adaptability.Finally, 17% of the reviewed papers include definitions and discussions of adaptability as both intentional and autonomous.This is often the case for review papers or theoretical papers that discuss several uses and references of adaptability.

Descriptive vs. normative
Taken together, explicitly and vaguely normative references and discussions of adaptability were almost twice as common as a descriptive framing (59% compared with 33% in Fig. 2B).Only 8% of the papers used descriptive as well as (vaguely) normative framings.The combination of framings is illustrated in Fig. 2C.

Categories of adaptability
When combining the two analytical dimensions, six types of references to adaptability emerge (see Table 3).A categorization of each individual paper is found in the Appendix (Tables A2 and  A3).In the following presentation of results, we merge the categories of normative and vaguely normative and provide illustrations and discussion of the four major categories: Autonomous-Descriptive, Autonomous-Normative, Intentional-Descriptive, and Intentional-Normative.Each category is illustrated by a few citations.Fig. 3 shows the number of papers per year and the proportions of the four categories.

Autonomous and descriptive
A combination of descriptive and autonomous framings were found in 17% of the papers (Fig. 2C).This is a classical natural science analysis of SES as CAS, without discussions of adaptability in relation to intentions and strategies, and without making explicit any system state as more or less desirable.Two citations demonstrate typical contexts.The first involves decisions based on modeling, not real actors, and we refer to this as autonomous (see Methods above).The second citation describes income as a factor determining adaptability without any desired norm.
In adaptive management, chosen strategies are seen as experiments that provide information about the system that is being managed, which can be used to refine future strategies [...] Every time an agent makes a decision, it is based on past and learned experiences and is, therefore, adaptive.(Bodin and Norberg 2005:175-177).
Households with a higher and more diverse endowment of these different forms of capital are more capable of coping with perturbations and adapting to change.[...] This diversity of income sources is an adaptive response to variable and unpredictable biophysical and socioeconomic environments, but despite this many households in rural areas are chronically poor.(Vetter 2009:32)

Autonomous and normative
This combination is indeed the most common, appearing in 33% of reviewed material (Fig. 2C).As mentioned before, actors and their intentions or strategies may be discussed in these papers, but if so, such discussions are not related to the adaptability concept.
The first citation below combines CAS with a vaguely normative (best fit) outcome.The second uses adaptive capacity vaguely normatively yet includes no reference to intentions or strategies.
The third adopts an explicitly normative view but offers no information on how adaptability is self-organized.Evaluation of parameters that indicate the system's adaptive capacity could, at least, be employed to check that proposed risk treatments are unlikely to reduce system resilience.The adaptive capacity of the system could also be used to provide an indication of its ability to respond to unforeseen impacts and disturbances.Adaptive capacity, therefore, becomes a measure of system performance that improves risk management and that could be included in the suite of performance measures used to evaluate sustainability.(Blackmore and Plant 2008:231) Interstate and national initiatives have established a watertrading system coupled with a system of catchment-scale regulatory water plans.If this system proves to be moreor-less self-organizing, encourages innovation and adaptation, and reduces salinity and water-table rise, it should enhance resilience.(Walker et al. 2009:20) Intentional and descriptive Some papers discuss adaptability in relation to actors' intentions and/or strategies but do not connect this to normative or prescriptive statements.This combination accounts for only 12% of the papers.It is common for descriptions of adaptive comanagement processes.Below, we cite two rather conceptual papers.The first citation describes a clash of framing and finds the normative definitions within adaptive governance inadequate.The second emphasizes intentions and politics and provides no normative meaning of adaptability.

These [community-based conservation] institutions can be conceptualized as complex adaptive systems because they are composed of interacting agents, have emergent properties resulting from agent interactions, can
Rather than the implementation of singular plans, adaptive governance emphasises the interaction of multiple institutions in guiding a complex system towards some more favourable state or trajectory (transformability) or maintaining it in a desired state or trajectory (resilience) (Walker et al. 2006a).Accepting that the outcomes of intervention will remain uncertain, adaptive governance emphasises flexibility, experimentation, and learning as strategies for anticipating and dealing with unintended consequences.Such governance approaches are thus deemed appropriate to situations of rapid change and high uncertainty.Nevertheless, they tend to assume that there are shared goals around what system properties should remain resilient, or that consensus can be built through the governance process.In this respect, adaptive governance is inadequate to deal with the kinds of clash of framing and value that emerge in the case studies I have described.(Leach 2008(Leach :1791) ) ...adaptation is not a predetermined outcome that arises deterministically from biophysical considerations.It

Intentional and normative
In this set of papers (18% of the papers analyzed), the use of adaptability recognizes the need to study people as strategically planning and intentional actors and also gives a norm on what is considered successful adaptive behavior.The key challenge is often described as mainstreaming of lessons learned, or avoiding socially defined negative outcomes.Unlike the normativeautonomous framing, the normative-intentional combination needs not portray adaptability as something positive.The first two citations below provide explicitly normative frameworks, suggesting that intentional adaptations and strategies of local actors are successful.The third suggests that actors can have incentives to adapt, or even increase their adaptive capacity, in a way that decreases social wellbeing (economic efficiency).
In a social-ecological system with high adaptability, the actors have the capacity to sustain the system in desired states in response to changing conditions and disturbance events [...] The second section describes the development of [Ecomuseum Kristianstads Vattenrike] and the selforganization process toward an adaptive co-management system [...] how one local individual played a critical role in leading change and transforming governance into an adaptive co-management system.(Olsson et al. 2004b:2-4) ... because human actions dominate social-ecological systems, the adaptability of such systems is mainly a function of the individuals and groups managing them.Their actions influence resilience, either intentionally or unintentionally (Berkes et al. 2003).Their capacity to manage resilience with intent determines whether they can successfully avoid crossing into an undesirable system regime or succeed in crossing into a desirable one.(Walker et al. 2006b:3)

In both responses involving rule-breaking-on the part of the Forest Department and local people-perverse
Ecology and Society 22(1): 12 http://www.ecologyandsociety.org/vol22/iss1/art12/learning results in a positive outcome with respect to the third resilience characteristic, the capacity to learn and adapt [breaking rules].However, it should be emphasized that the learning is perverse because it benefits few at the expense of many and endangers the forest resource.(Bingeman et al. 2004:111)

Power issues and transformations
Of the SES resilience literature published in the period 2001-2015, 44 out of 798 papers (less than 6%) mentioned "power" in the title, abstract, or key words, whereas 198 papers (roughly one in four) mentioned "transform*."Fig. 4 shows the total number of papers per year.It is clear that transformation and power issues are increasingly being addressed in SES literature, at least in absolute terms.

DISCUSSION
Our results show that the literature on resilience of SES is very diverse when it comes to how adaptability is defined, analyzed, and discussed.Previous studies have concluded that many strands of scholarship discuss adaptive capacity (not only the field of SES research), and they do so differently (e.g., Plummer and Armitage 2010:7-8).
Our results put numbers on this diversity (Fig. 2): 31% of the assessed papers do address actors with intentions and strategies.An additional 17% use both intentional and autonomous framings.Thus, we must reject the first hypothesis, that the SES literature treats adaptability as a capacity of a system to selforganize in an autonomous way without describing agency.Similarly, we find that 33% of the assessed papers use a descriptive framing, and an additional 8% have several framings.Although a normative use dominates, our results do not support a generalization that adaptability is used normatively in this literature.We can, therefore, also reject the second hypothesis, that SES literature necessarily makes normative judgements concerning what are "desirable" adaptations.

Autonomous or intentional dynamics
As discussed in the theory section, self-organization may refer to either autonomous or intentional processes.The former is inspired by CAS theory and sees self-organization as an uncoordinated process of adaptations resulting in emerging system properties at higher levels of social organization.This is common in the ecosystem resilience literature but, contrary to assertions by Olsson et al. (2015), the SES literature often describes self-organization as voluntary, intentional decentralized actions.
The papers conceptualizing adaptability as only autonomous (51%) did not assess peoples' strategies or social-ecological feedbacks, including learning or organizational/policy change in the context of adaptability, and therefore, did not really address the capacity to adapt.However, these papers seem to have a different purpose, so it may not be accurate to portray them as "failing" to recognize the organizational/political aspects of adaptations.Our impression from the assessed papers is that the analytical focus under study influenced the construction of adaptability: adaptations that appear to be autonomously selforganized in analyses at higher levels of abstraction or when assessing only ecological state and trends (MA 2005a), are instead described as driven by actors with (conflicting) intentions and strategies if another analytical focus is chosen.For example, Walker et al. (2004:1) use an autonomous CAS framing "to diagnose known examples of regional development" like the Kristianstad case, whereas local empirical analyses of the same case focus on intentional actors, networking, and institutions (Olsson et al. 2004b, Hahn et al. 2006, 2008, Johannessen and Hahn 2012).
The CAS framing in much of this literature has other advantages also in the adaptability discourse: in particular, in analyzing the "need" for adaptability in management of SES and types of capacities that will be required if crises cascade across scales and systems.Complex adaptive systems and other systems approaches in SES research enable analysis of anticipated vulnerability related to thresholds, tipping points, and ecological regime shifts (Walker et al. 2009).Such anticipated vulnerabilities are often analyzed within a (vaguely) normative context, assuming that adaptability is good without discussing agency.Although discussing vulnerability reduction in a normative framework may not necessarily be a problem-people can often agree that avoiding famine and drought is good (Duit et al. 2010-there are cases where famine and food provision are used as political tools (Keller 1992), and in such cases, an autonomous-normative framework is of course not very useful.
There is a huge leap from an analysis of adaptability as "needed" to an analysis of how it is organized by real people (Berkes 2009); whether or not related to a certain desired outcome.Recent literature on social learning and adaptive comanagement opens up the "black box" of self-organization and acknowledges that learning is not value free and that power issues are important parts of the system dynamics in a SES that sometimes block "desirable" adaptations and transformations (e.g., Plummer and Armitage 2007, Armitage et al. 2008, Berkes 2009, Reed et al. 2010, Österblom and Folke 2013, Nykvist 2014).This also holds for social-ecological memories as contributing to adaptability and resilience (Nykvist and von Heland 2014).
Identifying the need for adaptation is one research endeavor, and assessing the capacity to adapt, or the art of adaptation, is another.There seems to be a division of labor between researchers http://www.ecologyandsociety.org/vol22/iss1/art12/with different foci, just like Brand and Jax (2007:10-11) suggest a division of labor between a descriptive use of resilience in ecology (ecosystem resilience) and a "vague and malleable" use of resilience (social-ecological resilience) in transdisciplinary research (see also Folke 2006).Therefore, a transition from ecosystem resilience to social-ecological resilience (the focus for this paper), where adaptability is a part of resilience, can be expected to have some implications for the conceptualization of adaptability.The old (Garrett Hardin) framing of the "Tragedy of the Commons" used a maximum sustainable yield rather than a CAS framing, but like CAS, it tended to overlook the role of intention, communication, and collaboration.Hardin's analysis was "based on an extremely sparse view of the commons" (Ostrom 2007:15183).
Interestingly, the SES literature sometimes combines a CAS framing with agency by opening up the self-organization box and filling it with real people and political action.In the theory section, we exemplified this by the emergence of global governance to combat illegal fishing (Österblom and Folke 2013).In a second example, in an evaluation of adaptive comanagement, Plummer and Armitage (2007:65) used CAS and "the lens of resilience to [analyze] the role of institutions and power."Third, Galaz et al. (2010:371) used CAS theory to assess cascading ecological crises, arguing that the "policy failures that occurred were often not inevitable."Inside the self-organization box of adaptation, they discussed a combination of psychological, bureau-organizational, and political factors that decreased the adaptive capacity.Finally, the panarchy multiscale version of the adaptive cycle (Holling et al. 2002) has been employed by Hahn (2011) to analyze how the social memory at higher institutional levels was used strategically by some actors to change power structures and further develop the biosphere trajectory at the lower municipal level of Kristianstad.

Normative or descriptive positions
If science is what scientists do, we may conclude that "adaptability" is primarily a normative term in the context of resilience and SES.One-third of the assessed papers used adaptability in an explicitly normative way.Another third used a vaguely normative framing (Fig. 2).However, as 33% of the papers in this review suggest, it is of course possible to make a descriptive analysis of adaptability, just as democracy and other concepts with normative connotations can be analyzed descriptively.The fact that "maladaptation" is normative (negative) does not necessarily make "adaptation" normative (positive).Although "adaptation" is a more descriptive term in the literature, reflecting its original ecological use (Plummer and Armitage 2010:7), "having a capacity to adapt" appears to have normative connotations, e.g., in relation to capacity building.
A high capacity or ability to adapt is, of course, related to an expected increase in human wellbeing.Still, some caution is warranted.A high adaptability in multistakeholder situations characterized by genuine uncertainty, wicked problems (Ludwig 2001), or postnormal science (Funtowicz and Ravetz 1992) may result in lowering human wellbeing ex post.And perverse learning, perverse incentives, and economic rent-seeking may also coexist with high adaptability and result in lower human wellbeing overall (see citation above by Bingeman et al. 2004).Thus, adaptability should not be equated with a sustainable, or any other, outcome.Similarly, stakeholder participation cannot be assumed to be a suitable and/or sufficient method to achieve social learning in the sense of deeper learning among a range of actors (Elbakidze et al. 2013, Nykvist 2014).And social learning in turn should not be equated with proenvironmental or sustainable behavior or other desirable outcome (Reed et al. 2010).
As mentioned before, even if there is normative agreement on goals, such as the UN's announcement of its Sustainable Development Goal as the new global consensus, we should expect conflicting interests when policies are formulated for adaptations and transformations to meet these goals.Words like "desirable" should either be avoided in SES resilience literature or be conditional on the identification of an explicit normative framework or specific actors.Following this reasoning, adaptability could be defined as the capacity of actors to change ecosystem management and thereby avoid what they regard as undesirable regime shifts.Transformability could be defined nonnormatively as the capacity to break path dependency and shift toward a new development trajectory justified by a fundamentally different narrative.

Pluralism in social science
We found little support for associating the SES literature to the conservative social equilibrium ideas of sociological functionalism and its "inability to explain rapid social change" (Olsson et al. 2015:5).In fact, during the last decades, systems approaches like complexity theory have provided diversity and promising approaches in social science (Byrne 1998, Urry 2003, Castellani and Hafferty 2009).For example, Schwandt and Szabla (2003) identify a conceptual congruence between Giddens' structure and agency theory and complexity theory, and CAS has also inspired anthropology (Lansing 2003).
Our review of the 183 papers indicates that complexity theory and CAS can be applied to the issue of adaptation in many different ways; our categorization shows that it provides pluralism.Pluralism is ironically what Olsson et al. (2015:9) also advocate while they dismiss the resilience approach, which they argue is "rooted in complexity theory" and functionalism and therefore "becomes the equivalent of stability and harmony" (Olsson et al. 2015:5).
The alleged assumption of harmonious consensus building requires reflection on the intentional-normative framing: that people are described as having intentionality and seeking common interests.An insight from social science, prominent in the critiques of resilience, is that harmony and consensus can never be assumed, not even on knowledge.Adaptability of SES has been defined as the capacity to adapt and respond to change "in an informed manner" (Folke 2006).This raises the question of whose knowledge should count, or which knowledge systems (local, indigenous, or scientific) and strategies qualify as more or less "informed." Consider two adaptive strategies proposed to feed the poor and achieve the second Sustainable Development Goal (Zero Hunger): agroecology and genetically modified organisms (GMO).Pretty et al. (2006) and an international report (United Nations 2011) have suggested agroecology as a promising strategy to adapt or transform agriculture in low-income countries to enhance yields sustainably.Other actors have suggested input http://www.ecologyandsociety.org/vol22/iss1/art12/subsidies to GMOs and another "green revolution" for the same purpose (Huang et al. 2002, Carpenter 2010).These two different strategies reflect different worldviews and priorities held by different stakeholders.They are both supported by scientific research and thus ecologically "informed."They are both "desirable" within their respective political and scientific frameworks, and from an analytical perspective, both manifest adaptive capacity.
The ultimate normative challenge concerns claims for efficiency, which means connecting a particular adaptation or transformation strategy to desirable outcomes.The agricultural example above suggests that both ex ante and ex post evaluations of such strategies (e.g., what is efficient and desirable for whom?) sometimes belong to the political struggle of establishing the (scientific) truth.

CONCLUSION
We conclude by emphasizing the plurality of approaches in the research on resilience and SES.There are signs that ontological clashes between social sciences and SES research have resulted in constructive pathways for dealing with questions of adaptation and adaptability, drawing on core concepts and insights "from both sides."Rather than assuming consensus and selforganization, resilience in SES has been employed empirically to explain how social learning and networking are used to alter power structures and achieve adaptations and transformations.We believe this constructive debate needs to continue to advance science and provide adequate policy support for sustainability transformations.
Responses to this article can be read online at: http://www.ecologyandsociety.org/issues/responses.php/9026 No reference to intentions or strategies.Loops of learning are assumed to emerge to comanagement.

S
No particular system state or desirable end is referred to.The process of adaptive comanagement is discussed descriptively.No discussion of strategies of actors, all events are just unfolding.p. 1083.Resilience is not a normative term; system configurations characterized as resilient may be either desirable or undesirable.In particular, resilience theorists are interested in understanding where resilience, adaptive capacity, and the potential for innovation reside in linked social-ecological systems and how these attributes can be gained, lost, or preserved (Walker et al., 2002).p. 1083.In this type of trap, which has been referred to as the rigidity trap by resilience theorists (Gunderson and Holling, 2002;Allison and Hobbes, 2004;Atwell et al., 2009b), the high adaptive potential and connectedness of social actors makes it possible to continue to invest in the current way of doing agriculture, in spite of the mounting social and ecological deficits and economic inefficiencies (Harvey, 2004), associated with this trajectory.p. 1087.Our participants indicated that regional coordination of scientific monitoring, adaptive management, and enforceable environmental standards are foundational to long term,multi-objective change given the complexity of the Corn Belt system.While such an approach was seen by our workshop participants as having the potential to link agricultural and conservation objectives, they also

S
p. 9. Tongass system became rigid and maladaptive in response to a confluence of events that fostered simultaneous collapse dynamics in the policy and economic subsystems.Within 5 years of the passage of the TTRA, Tongass timber production declined by roughly 85% and stabilized at an output level equivalent to the pre-industrial production level (Fig. 1).p. 11.The preceding narrative suggests the importance of changes in federal policy and larger-scale economic factors as the key drivers of adaptive-cycle dynamics in Tongass governance.Policy mobilized much of the initial growth, provided much of the stability during the conservation phase, and served to destabilize-at first incrementally and then rather suddenly-the industrial forestry regime of the Tongass.p. 13.In the current state, Tongass managers are pitted between environmental advocacy groups (via litigation and appeals) and pro-timber legislators and executive officials (via budgets and national leadership).This situation acts to maintain day-to-day Tongass decision making in a highly stable but inflexible state that constrains managers from responding adaptively to changing conditions and, ultimately, from finding a sustainable direction for the future governance of the Tongass.As long as the dominant venues of stakeholder input in Tongass decision making are adversarial-i.e., through appeals and litigation-the institutional subsystem will remain "trapped" in the collapse p. 6.There is a systematic ambiguity in the term "adaptive," as used in "adaptive cycle."The key question is: What adapts?In Holland (1992:184-185), adaptation is a property of organisms or of analogous components of an artificial system.p. 7. The panarchical perspective has also had little to say about the long-term growth of world population, energy use and polity size, and the power of technology available for human use.As with the comparative neglect of conflict and elite dynamics, this is perhaps unsurprising, given its disciplinary roots in economics, ecosystem science, institutional research, and adaptive complex system theory (Holling 2003).However, all these factors have enormous implications for the past and future development of social-ecological systems at all scales.
Several perspectives, review of adaptive cycle adaptations therein.Notes that the adaptive cycle critically does not address power and political dimensions.

B
No particular norm.Theoretical discussion on adaptations only.Janssen et al. (2007) p. 309.systems subjected to a particular type and degree of variability may become highly optimized to tolerate this variability (this characteristic of adaptive systems is referred to as highly optimized tolerance or HOT).In so doing, how ever, the system may become more brittle and susceptible to changes that may occur in the type and degree of variability to which it has become highly adapted or to new types of disturbances.p. 311.We may distinguish two classes of adaptations.First, people have developed institutions over time, intentionally or not, to spread resource-use intensity over space and/or time in accordance with particular variability regimes.p. 312.The second class of adaptations is characterized by those directed at managing discrete disturbances like droughts, cyclones, and price fluctuations.p. 312.For example, from the perspective of higher level authorities, local SES might not be recognized as being well adapted to the challenges it faces The ambitions of higher level authorities may lead to changes in local institutions in the expectation of meeting their goals.Lack of understanding of the SES leads to an inability to meet participants' goals and a reduction of the performance of local and sometimes larger scale SESs.

D
Adaptations  1997;Arthur, 1999;Kooiman, 2003;Teisman et al., 2009).p 364.From field work across the world, the 23 case studies reveal a rich variety of circumstances ranging from environmental emergency migrants, flooding and resettlement suggesting an analytical distinction between rapid-and slow-onset events.She notes that the development community often characterizesmigration as a failure of adaptation, rather than as a formof adaptation to environmental and climate change.p 365.Such thorny normative and conceptual issues notwithstanding, resilience thinking also holds a great deal of potential for renewing thewider governance research agenda.In particular, it invites us to consider fundamental issues of change and stability, adaptation and design, hierarchy and self-organization in the study of multilevel governance systems.Moreover, in addition to ''tradi-tional'' benchmarks such as Review CAS as applied to SES, but also discusses problem with spontaneous view and recognise that adaptations are always towards a norm.
Both Discusses whether migration is a failure of adaptation.Vulnerability resulting from maladaptation is part of normative evaluation.But adaptation is also discussed in terms of adaptive agents responding to impulses.SP efficacy, accountability, and equity usedwhen assessing public governance, a resilience perspective on governance would also consider issues of human-environmental interactions, vulnerability resulting from mal-adaptations, and innovation capacity as integral parts of evaluating a given governance system (Nelson et al., 2007)
Fig. 2. (a) Autonomous vs. intentional."Both" refers to discussions of adaptability where reference is made to both perspectives.(b) Descriptive vs. normative and vaguely normative."Several perspectives" means using descriptive as well as (vaguely) normative framings of adaptability.(c) Categories of adaptability.The combination autonomous and normative is most common framing, appearing in 33% of the categorized papers.

Fig. 3 .
Fig.3.Share of papers that refer to adaptability according to our four categories (left axis) and total number of papers (right axis).This analysis was based on the first bibliometric search (see Methods).

Fig. 4 .
Fig. 4. Total number of papers and number of papers that also mention power and transformation, respectively.
each of the three properties (potential, connectedness, and resilience) in the adaptive cycle is given two nominal levels, either low or high, then the adaptive cycle model uses only four of a possible eight combinations (23) of the three properties, and two of the other four combinations are suggested as pathological states, labeled the poverty trap and the rigidity trap byHolling et al. (2002c), which are departures from the adaptive cycle (Fig.3).

Table 1 .
Coding of research question 1, autonomous vs. intentional.Each row represents sufficient but not necessary conditions

Table 2 .
Coding of research question 2, descriptive vs. normative.Each row represents sufficient but not necessary conditions.

Table 3 .
The six possible combinations of framings.For exact coding of each framing, see Methods.
. 2: Finally, we discuss how their integration in adaptive co-management systems may provide more efficient management of biodiversity and ecosystem services in the NUP….investigate how adaptive capacity can be built to better respond to social-ecological change farmer bases his labor allocation decision on an evaluation of the returns per effort obtained from his farming and fishing activities in the previous year.Hence, he adapts his strategy by learning from past experiences to find the mixture of activities that yields the highest returns per effort.p 501 This learning process and its transient dynamics can lead to suboptimal outcomes when agents do not have enough time or capital to adapt their strategy to the local conditions before going bankrupt.Diversification of water use by balancing the needs of different water users as demonstrated here can contribute to an enhancement of adaptive capacity and thus resilience.
Berkes et al. (2003)p a social capacity for urban ecosystem management to respond to change, and to develop policy directions that can help build resilience to deal with further change.Berkes et al. (2003)refer to such a capacity as "adaptive capacity." p.16:A crucial part of building adaptive capacity is a governance system that can learn from experience and generate knowledge across organizational levels.p As a result, adaptive actions often reduce the vulnerability of those best placed to take advantage of governance institutions, rather than reduce the vulnerability of the marginalized, or the undervalued parts of the social-ecological system(Adger et al., 2005a).p. 277.Adaptation does not necessarily entail changes in system boundaries in order to build resilience.And in the same fashion, adaptation strategies that include radical change of resource use (in location, economics or significant land use change for example) may not necessarily be a symptom of a lack of resilience.
Not only can the natural world be analyzed as a complex dynamic system.It is also possible to view human-made governance systems consisting of institutions, networks, bureaucracies, and policies as examples of complex systems inwhich adaptive agents respond to external and internal impulses (cf.Jervis,

Table A2 .
Categorisation of all 183 papers.

Table A3 .
List of all 183 papers for the structured literature review.Molnar, G; Sarvari, A; Nagy, Z 2008 Assessing the Resilience of a River Management Regime: Informal Learning in a Shadow Network in the Tisza River Basin Ecological Resilience by Placing Science at the Decision Table: the Role of the San Pedro Basin (Arizona) Decision-