Behavioural diversity in fishing—Towards a next generation of fishery models

Decades of fisheries research and management have significantly advanced the understanding and management of fisheries by, for example, focusing on the temporal and spatial dynamics of fish stocks (Botsford et al., 2008; May, Beddington, Clark, Holt, & Laws, 1979; Pikitch et al., 2004), the complexity of food webs and ecological systems (McLeod & Leslie, 2009) and the role of regulations Received: 17 November 2019 | Revised: 10 April 2020 | Accepted: 16 April 2020 DOI: 10.1111/faf.12466

| 873 WIJERMANS Et Al. and economic incentives in shaping fishing practices (van Putten, Gorton, Fulton, & Thébaud, 2012). However, numerous fish stocks worldwide remain overharvested despite these advances in fisheries science and management. One important reason for the lack of progress is that fisheries management typically relies on models of human behaviour and assessments that are unable to account for social dynamics, and in particular for the diversity and adaptability of fisher behaviour (van Putten, Kulmala, et al., 2012;Wilen, Smith, Lockwood, & Botsford, 2002). Recent literature stresses how this omission is problematic: the limited understanding of the diversity of fisher behaviour may limit management interventions anticipation of fishers' response to regulation (Fulton, Smith, Smith, & van Putten, 2011;van Putten, Kulmala, et al., 2012;Salas & Gaertner, 2004). For instance, behavioural responses have been shown to play an important role in the success or failure of marine protected areas (MPAs). MPAs have swiftly become a widely used management tool based on their potential benefits to both people and planet (Chaigneau & Brown, 2016;Edgar et al., 2014). However, MPAs often fail to achieve their objectives: In the North Pacific Trawl fishery, for instance, fishers' adaptive responses to MPAs have caused a dramatic increase of prohibited bycatches of one species (Pacific Halibut (Hippoglossus stenolepis, Pleuronectidae) while protecting another (Red King Crab (Paralithodes camtschaticus, Lithodidae) (Abbott & Haynie, 2012). Furthermore, compliance with MPAs may be affected by perceptions of inequality or lack of fairness in the distribution of MPA benefits, which has led to poaching fish despite their generally positive attitudes towards the MPA and even despite receiving benefits from it (Chaigneau & Brown, 2016).
There are several reasons for why human behaviour is not represented more adequately in conventional fisheries science and management. Firstly, knowledge about human behaviour in general, and fisher behaviour in particular, is primarily held within the social (fishery) sciences and, with the exception of fisheries economics, is less visible in the fisheries literature which is dominated by natural science contributions. This lack of visibility means that many fisheries scientists, decision-makers and regulators remain unfamiliar with the contributions that social science knowledge, theories and methods can make in fisheries science and management (Heck, Stedman, & Gaden, 2015). Secondly, fisheries management is often supported by, and dependent upon, formal modelling (Hall-Arber, Pomeroy, & Conway, 2009). Such models rarely include fisher activity dynamics beyond economic considerations (van Putten, Kulmala, et al., 2012).
We answer these calls by highlighting key insights about fisher behaviour from the social sciences and by demonstrating how such insights can be modelled to support the understanding of the impact of behavioural diversity on fishery outcomes. As part of this effort, and to structure our study, we seek answers to four questions: 1. Which social science insights are relevant for understanding fishers' behavioural diversity? 2. How are aspects of fishers' behavioural diversity represented in recent fishery models?

How can such insights be integrated in a fishery model? And what
can we learn from the model regarding the implications of behavioural diversity for fishery sustainability and management? 4. What can we learn from this research for developing the next generation of fishery models?
This paper aims to contribute to the development and use of a new generation of fishery models that integrate social science insights to enhance our understanding of why and how fishers' behavioural diversity affects fishery sustainability and fishery management. Thus, we first review and synthesize insights from social scientific fishery studies and evaluate whether and how recent fishery models incorporate insights on fishers' behavioural diversity (Section 2). Section 3 presents an approach for incorporating these insights into an agent-based model (ABM). The model exemplifies the formalization of human (fishers') behavioural diversity in the form of fishing styles, thus reflecting both small-and large-scale fisheries (Boonstra & Hentati-Sundberg, 2016). The ABM is then used to investigate the implication of fishers' behavioural diversity on fishery outcomes (stock, profit, satisfaction) and effectiveness of management interventions (Section 4). We conclude with a discussion of our findings' implications for the development of next-generation fishery models (Section 5).

| FIS HER B EHAVI OUR IN FIS HERIE S SCIEN CE
The point has been made often enough to almost become a platitude: human behaviour is a key uncertainty for fisheries management, because it receives too little attention in conventional fisheries science and management (Fulton et al., 2011;van Putten, Kulmala, et al., 2012). Several excellent reviews of fishery models representing both fisher or fleet behaviour and fish population dynamics have recently proved this point (Fenichel et al., 2012;Girardin et al., 2017;Hamon, Frusher, Little, Thébaud, & Punt, 2014;van Putten, Kulmala, et al., 2012). They stress the need for a more realistic inclusion of human behaviour and, in particular, the dynamics of resource users.
Addressing this need crucially requires an overview of the available and relevant knowledge on human behaviour and its potential for illuminating fishery models.
In this section, we first outline key knowledge on fisher behaviour from social scientific studies as a first step towards a more realistic representation of human behaviour in fisheries modelling and management. Rather than attempting a futile exhaustive overview of all social sciences and humanities, we concentrate on literature that engages with fisher behaviour directly or provides central insight. Subsequently, we review a representative set of recent fishery models (Section 2.2) to evaluate how they consider knowledge on fisher behaviour.

| Understanding fisher behaviour: Contributions from social sciences
Several social science traditions highlight the importance of the diversity and variability of fisher behaviour (Boonstra & Hentati-Sundberg, 2016;Gustavsson, Riley, Morrissey, & Plater, 2017;Urquhart, Acott, Reed, & Courtney, 2011). With a focus on a classic distinction Boonstra, Björkvik, Haider, and Masterson (2016) between motivations (desires, aspirations, values, etc.), abilities (agency, power, perceptions, etc.) and context (interactions, institutions, social structures, social and natural environments, etc.), we present contributions from the social science literature on fishers' behavioural diversity through a discussion of motivations, abilities, livelihoods and social interactions.

Motivations
In fisheries science, and especially for bioeconomic models, scholars commonly assume that fishers weigh relative costs and benefits to realize their greatest personal (economic) gain (for the theoretic statement see Becker (1986) and Clark (2006b); for applications to fisheries see Sutinen and Anderson (1985); Anderson and Lee (1986)). The central and single motive identified is thus the desire to maximize gratification and to avoid punishment. While these assumptions can be readily applied in fisheries models and may predict outcomes, they poorly represent the empirical complexity, in particular the variety of fishers' motivations leading to the decision whether to go fishing or not.
Sociological and anthropological studies of fishers' behaviour, in contrast, typically assume motivational differences. They show that fisher(s)' practices and responses do not only derive from a desire to (deliberately and rationally) realize their greatest personal (economic) gain, but also from a desire to conform to social norms, to uphold morality and identity, to experience esteem (or avoid shame) and solidarity and, by doing so, to maintain fishing not only as a business but also as a way of life (Hall-Arber et al., 2009). This extended spectrum of fishers' motivations implies that a singular focus on (economic) interests and deliberate decision-making is insufficient.
The importance of considering more than this single (economic) motivation can be illustrated by the question of why fishers often continue to fish despite considerable social and ecological setbacks (e.g. Daw et al., 2012). From a purely economically reasoning perspective, this outcome would be difficult to explain. Sociological or anthropological explanations often refer to fishers' desire to maintain an occupational identity and culture, meaning that for many fishers, fishing is not merely a job but rather a lifestyle through which they self-identify (van Ginkel, 2005;Gustavsson et al., 2017;Pollnac & Poggie, 2008). Continuing to fish under adverse conditions can thus be explained as an effort to uphold a preferred self-image.

Abilities
Behavioural diversity can also be rooted in fishers' differentiated abilities, that is their various degrees of access to and control over economic, cultural and social capital. Social scientific studies have, for example, analysed differences in fishers' (ecological) knowledge and skills (Pálsson & Durrenberger, 1990), that is their repertoire of experiential and tacit knowledge and skills that fishers embody and which is (re)produced through their working in specific environmental and social contexts. Social scientific research conducted on this topic (see Hind (2015) for a comprehensive overview) assumes that fishers whose livelihoods directly depend on local ecosystems develop a rich and nuanced understanding of these (Johannes, Freeman, & Hamilton, 2000). Further studies document how different fishing practices and styles (re) produce various types of knowledge (e.g. Lauer & Aswani, 2018).
Moreover, studies point out how knowledge and skills-such as observing tides and weather patterns, maintaining safety at sea, finding fish, equipment handling, but also respecting other fishers, upholding norms and values related to catching fish-serve as cultural and social capital that fishers use for maintaining and developing their businesses and livelihoods (see e.g. Gustavsson et al., 2017).

Livelihoods
Social scientific studies of fishers stress that fishing practices need to be understood in relation to fishers' livelihood diversification (Blythe, Murray, & Flaherty, 2014;Byron, 1986;Coulthard & Britton, 2015;Salmi, 2005). Fishers' livelihoods are often sustained not only through fishing but involve a portfolio of income-generating activities such as farming, trading or industrial work, or they are supported through social welfare payments or subsidies. Further, these diverse activities are not only performed by fishers but in collaboration with their family members. Reflecting such a variety constituting fisher' livelihoods in research is important because the extent of non-fishing income-generating activities influences fishing effort.
On the one hand, non-fishing activities limit fishers' time available for fishing. On the other hand, revenue from non-fishing activities can be invested in developing the fishing enterprise. Alternative income sources can assist fishers in coping with periods of low fish stock levels or when fishing opportunities are limited due to, for example, policy changes.

Social interactions
Another recurring theme in the social scientific literature on fisheries is the various ways in which fishers socially connect with one another. They, for example, simultaneously compete and collaborate (Basurto, Blanco, Nenadovic, & Vollan, 2016;van Ginkel, 2005;Löfgren, 1972;Pollnac & Poggie, 1991). This means that, on the one hand, fishers are understood as individualists trying to gain and protect (knowledge of) lucrative fishing locations (Byron, 1986). On the other hand, they are also portrayed as members of close-knit family and friendship communities in which they collaborate and share knowledge and resources (Acheson, 1981). Some scholars view the performance of these different social roles as paradoxical (McGoodwin, 1991), while others point out that they do not necessarily contradict: fishers can be collaborators within the social groups they identify with and be competitive towards outsiders (Acheson, 1981;Pálsson, 1994). Moreover, the literature indicates that such social dynamics can change over time. Management arrangements, for example individual transferable quota, have changed the competition between fishers from a race for fish to a race for capital and quota (Acheson, Apollonio, & Wilson, 2015) which brings us to the final aspect of social interactions. Different regimes of fisheries governance (such as state-led, market-oriented or community-based arrangements) impact fisher behaviour (Acheson, 2006;McEvoy, 1986;Ostrom, 1990). A major concern for social scientists studying marine governance is to understand how such institutions shape dependency and power relations between fishers, communities, markets, governments and marine environments (Boonstra & Österblom, 2014).
In the subsequent section, we consider the extent to which these aspects of fishers' behavioural diversity-variety of motivations; abilities; livelihoods; and social interactions-have been incorporated into fisheries models.

| Diversity of fisher behaviour in fishery models
To assess the inclusion of aspects of fishers' behavioural diversity in dynamic models of fisheries, we reviewed 29 recent fishery models that incorporate endogenous and dynamic fisher behaviour. The selected articles were derived from filtering 1,290 articles found in the "Scopus" database published after 1999, that is in the last 20 years.
The results were filtered to include dynamic, formal model papers that were not prescriptive, that is focusing on how fishing behaviour should be (see Appendix S1).
The review substantiates the claims that current fishery models pay too little attention to the diversity of human behaviour (Fulton et al., 2011;Hall-Arber et al., 2009), but also highlights progress in the models that do. We evaluate whether models reflect behavioural diversity, in particular a variety of motivations, abilities, livelihoods and social interactions; and if so, whether they are based on social (fishery) science insight. We selected only fishery models that explicitly considered social and ecological interactions, as well as their respective internal dynamics. Consequently, we excluded models that generate diverse behavioural outcomes without including behavioural diversity (e.g. Chakravorty & Nemoto, 2000;Metcalf, Moyle, & Gaughan, 2010), when behavioural diversity in models was produced through stochasticity in a deterministic model or when, for example, interactions with spatial and seasonal heterogeneities in the distribution of natural resources result in behavioural diversity (Gasche, Mahévas, & Marchal, 2013). Lastly, we did not restrict our selection of models in terms of their functional form (e.g. equations in bioeconomic models) or type of fishery (e.g. commercial fisheries). Within our selection, we identified two different categories of model foci.
The first category represents fishers' behaviour endogenously, that is fishers may adapt their behaviour when their own state or the state of the environment changes, and is concerned with fishery outcomes from fisher-fish stock interactions, that is studying the link between fishers' behaviour and its effect on the fish stock and vice versa, or with the effect of management on fishery outcomes.
Fishing actors in these models represent individual fishers (e.g.  (Hunt et al., 2011); the inclusion of fishers able to switch strategies (Bischi, Lamantia, & Radi, 2013;Brede & de Vries, 2010); and the effect of fishers' ability to learn to refrain from competitive behaviour (e.g.

| Behavioural diversity represented in the reviewed models
We are interested in whether and how fishers' behavioural diversity is reflected in our selection of fishery models. A comparison (Table 1) shows that none of the reviewed models reflect all four aspects of behavioural diversity that we highlighted above (Section 2.1). Actors' decision-making is mostly reflected by assumptions underlying rational actor theory, that is, fishers acting as firms, maximizing their profit, such as expected utility theory or random utility theory, mainly referring to economic factors (see also van Putten, Kulmala, et al., 2012). For instance, the fishers/vessels/fleet in most models (21/29) has a single economic motive: maximizing/satisficing profit or catch.
Consequently, other factors of influence are directly or indirectly economic, for example expected catch, costs of effort/travel, prices and regulations. While the behavioural focus differs slightly among models, there is little variety within the individual models. Regarding the behavioural diversity actors display, the models typically focus on one type of behavioural choice, for example effort allocation. The selected models that do consider behavioural diversity are discussed in more detail below (see also  (2010) demonstrate that a community-oriented long-time horizon harvesting behaviour is useful because it leads to a relief in pressure on the resource, but also to smaller fluctuations in the fish stock, thereby reducing the risk of overharvesting. They also find that an overharvested resource situation favours short-term and individualistic behaviour.

Brede and de Vries
Heterogeneity in abilities is the most common way behavioural diversity was addressed in the models reviewed. Ability reflects fisher characteristics that, for example, include how well a fisher can find or catch fish and/or is knowledgeable of when and where to fish. Heterogeneity is further commonly represented by making differentiations between fishers or vessels in terms of, for example, business strategy, ability to catch fish and the cost of fishing (Merino et al., 2007); their effect on catch, CO 2 impact, economic performance and job availability (Sigurðardóttir et al., 2014); or the minimum functioning crew size, maximum trip duration and speed (Pelletier et al., 2009). These models generally study the consequences of integrating heterogeneity in, for example, strategy or gear to account for behavioural response to policy. In sum, these

| MODELLING FIS HER S' B EHAVIOUR AL DIVERSIT Y
We developed an ABM to represent the four dimensions of fishers'  . The stylized model is thus sensitive to conditions in particular contexts, but it is not specific to a particular context.
The purpose of the model is then to: (a) demonstrate a formalization of fishers' behavioural diversity based on social science insights (Section 3.1); and to (b) demonstrate the insights obtainable through using the model as a virtual laboratory for exploring the implications of this diversity when seeking to achieve sustainable fisheries (Section 4).

| Formalizing the biophysical environment
FIsher BEhaviour model includes the representation of a biophysical environment (the sea and the fish). The sea reflects a space with various fishing grounds that differ in their fish abundance and remoteness. It is represented as a grid (50 × 56), where each grid cell (patch) represents a fish stock. Patches are grouped into four regions which correspond to the fishing grounds of the Swedish Baltic Sea fishers from which the fishing styles typology was derived. The regions vary by distance from the home port of the fisher: region A is close to the coast; region B is further out; and regions C and D are far offshore.
A patch can sustain more or less fish, reflecting spatial differences in carrying capacity. Patches with high carrying capacity are considered "hotspots." All fish populations have the same growth rate, and the fish do not move between patches. Fish population growth is represented by a standard discrete logistic growth model with growth rate and carrying capacity (Clark, 2006a;Schaefer, 1954).
Harvest levels are determined by the group of fishers.

| Formalizing fisher's behavioural diversity
The fishers in FIBE are represented by agents whose behavioural diversity is modelled according to the fishing styles. Each fisher agent reflects an individual with their own experience and interaction with the social and biophysical environment, while always being an instantiation of one of the three fishing styles: the archipelago, coastal or offshore trawler fishing styles ( Table 2 details  For the representation of the fishing styles, this involved distilling and formalizing the key elements and processes of fishers' behavioural diversity (see Figure 1 and Table 2 All fisher agents in the model face two main choices: (a) to fish or not to fish; and if they go, fishing (b) where to go fishing. To design these choices, a combination of empirical details and theoretical assumptions about human behaviour, preferences and choice was made, for example bounded rationality operationalized as limited memory. This is common practise as descriptive forms of knowledge are rarely specific enough for computational representation (Sawyer, 2004;Schlüter et al., 2017).

2.
A coastal fisher agent decides to go fishing when the trade-off between expected profit and time not spent at home is not too big: if a coastal fisher agent has satisfied its preference for staying at home and expects a profit, it will go out and fish. Coastal fisher agents stay at home when staying home preference is not satisfied or when they do not expect a profit.
3. An archipelago fisher agent goes out to fish when it needs to, that is when it has not caught enough in the previous week or if it has negative financial capital. In addition, if the archipelago fisher TA B L E 2 Description of the three fishing styles that are modelled in FIsher BEhaviour model

Fishing style Description
Archipelago General: Swedish Baltic archipelago fishing is a very particular and traditional style of fishing, requiring substantial investments in learning and in material resources Practice: Fishers localize fish close to harbours with passive gear. Traditionally (50-80 years ago), they target a broad portfolio of freshwater and coastal species. Recently, this has become increasingly restricted due to management regulations and loss of infrastructure Motivation: Fishers emphasize the importance of being able to catch multiple species and being self-reliant, that is avoid becoming too dependent on banks for investments

Coastal
General: This style represents the dominant type of fishery in the Baltic Sea. It forms an intermediate between the other two styles in terms of scale of operation and motivation Practice: Fishers combine passive and active gear and fish in coastal areas. Atlantic herring (Clupea harengus, Clupeidae), Atlantic cod (Gadus morhua, Gadidae) and Atlantic salmon (Salmo salar, Salmonidae) are the dominant species targeted in this style. Of these three species, cod is currently the dominant target species since fishing herring with passive gears is not profitable and offshore salmon fishing is prohibited Motivation: Similar to the archipelago style, but with a blend of entrepreneurial spirit (see description of offshore trawling style).
Fishers also emphasize the importance of being able to catch multiple species and being self-reliant, that is avoid becoming too dependent on banks for investments

Offshore trawling
General: This fishing style represents specialized fishing using trawler ships (both pair and single trawls) with modern fish-finding and fish-processing technology Practice: Fishers have mobile gear (trawling) used on large vessels (often larger than 20 m). The target species are Atlantic cod, sprat (Sprattus sprattus, Clupeidae) and Atlantic herring (Clupea harengus, Clupeidae) Motivation: Fishers have a strong entrepreneurial spirit in fishing. They are competitive, seize opportunities, invest and are willing to take risk when there is a profit to be made thinks fish is scarce, it can decide against fishing and instead reduce living expenses.
While the trawler fisher agents' fishing activities are independent of the weather, both the archipelago and coastal fishers do not go out to fish when the (stochastically determined) weather is bad.
We connected the reasoning underpinning each fishing style to the way the behaviour selection process works, that is maximizing OR the fisher's decision will either be informed by its own past experience (memory of good spots) or by what others do (social influence).

| E XPLORING THE ROLE OF FIS HER S' B EHAVIOUR AL DIVER S IT Y
To illustrate the kinds of insights obtainable from a model that reflects fishers' behavioural diversity, this section highlights two kinds: firstly, system-level understanding-which is the more traditional use of models-to present emergent patterns of each fishing style at the system level (the sustainability of a fishery) and secondly, an understanding that stems from connecting system-level outcomes to the underlying mechanisms (how the actions and interactions of individual fishers bring about the system-level outcomes) to explain the emergence of system-level patterns. MSY is a biological measure that indicates the stock size at which the reproductive rate of the population and hence the harvestable surplus is highest. It is a key indicator used in contemporary fishery management, despite some critique (e.g. Hilborn, 2004;Larkin, 1977). We call this benchmark "theoretical optimum management." For the benchmark calculations, we use a single fish stock and assume that fishers follow an optimum fishing strategy, as commonly assumed in bioeconomic fishery models. We calculate the benchmark MSY profit as the total profit that can be made in this fishery assuming fishing style-specific costs. This benchmark serves to assess the relative differences in fishery outcomes among the different fishing styles. We assess three types of fishery outcomes: the fish stock size (ecological), the profit (economic) and fishers' goal satisfaction (social). The first two are assessed relative to the stock size and profit of the theoretical optimum management benchmark.
Both the experimental design and benchmark are described in more detail in Appendix B4 in Appendix S2.

| Experiment 1: Effect of fisher behavioural diversity on fishery sustainability
This experiment demonstrates that the behavioural assumptions underpinning the different fishing styles have a strong effect on fish stock levels, profits and fisher satisfaction (Figure 4). The fishery F I G U R E 1 Conceptual representation of the fishing styles with key attributes and decision models. Note that bounded rationality is formalized by agents not being able to anticipate the probabilities and consequences of their actions (for all styles) F I G U R E 2 Decision trees for the choice whether to go fishing for each fishing style | 881 outcomes of the formalized archipelago, coastal and trawler fishing styles differ significantly from the benchmark of theoretical optimal management (the dotted line at 1.0 on the y-axis, Figure 4).
Of the formalized fishing styles, the archipelago fisher agents under-exploit, while the coastal and trawler fisher agents overexploit and deplete the fish stock (albeit at different speeds). Differences

F I G U R E 3
Decision tree for the choice where to fish, integrating the fishing styles

F I G U R E 4
The effect of diverse fisheries on the stock, economic profit and fishers' satisfaction. Each line (in each graph) represents the mean outcome for one fishing style (see legend) over 1,000 repetitions of the simulation experiments. Benchmark scenario: the thin dotted line in stock and profit graph. Note: In these experiments, there is no interaction between fishing styles agents, that is they do not compete for the same resources

| Experiment 2: Effect of a policy intervention on fishery sustainability
To explore the effect of a policy intervention for the different fishing styles, we performed the same experiment as above but introduced a fuel subsidy which reduces travel costs for all three fisher styles. The fuel subsidy has no or only a detrimental effect on the resource stock (see Figure 5). The policy affects fishers' profit, but the resource stock gains no observable benefit. The policy actually proves detrimental for the fish stock for the coastal fishing style agents. The financial support delays coastal fishers' realization that they have to reduce their catch.
The fuel subsidy thus appears to enable coastal fishers to catch more while being in a more resource-scarce situation, resulting in increased overexploitation. Interestingly, even though the profit is higher due to the policy intervention, the coastal fishers are less satisfied. The choice to fish (or not) by both the archipelago and trawler fisher agents is not affected, that is they do not go out less or more frequently and they merely have more profits when catches are large enough.

| Multilevel understanding
In these experiments, we analysed system-level patterns (stock size, profit and satisfaction) and highlighted differences between the fishing styles (trawler, coastal and archipelago). Some results may appear intuitive, but others leave questions that require closer Likewise, the outcomes for the trawler fishery can be understood by looking at the way its decision-making is characterized, in particular the motivation that underpins the behaviour of trawler fisher agents in relation to the social-ecological dynamics. A trawler fisher agent is motivated to maximize profits which drive the trawler fisher agents to go out and fish as much as possible for as long as they expect a profit. The model shows how under these assumptions the trawler fisher agents depletes the fish stock faster than they perceive signals from the declining fish stock. They are thus unable to respond in time and reduce fishing. Their delayed response to catch fluctuations is due to not having full information, that is bounded rationality. Their behavioural choice is based on past experiences of having found fish and earning a profit. A few days of bad catches do not immediately change this expectation, which explains their tendency to overfish.
For the coastal fishery outcomes, we ran some additional experiments to better understand the results. When following the reasoning of the decision-making type, a coastal fisher agent is similar to a trawler, with one major difference which is that they value other things than merely profit maximization: they also value spending time at home. When looking at the social-ecological situation, the coastal fisher agents find themselves in a different situation than the trawlers: they notice that they are overfishing (i.e. decrease in profit and satisfaction) and adapted their behaviour. Since the overfishing was not as extreme (or fast) they could avert a fish stock collapse.
But questions remain: for example, why do their catches reduce?
Do they go out fishing less often? Or do they not find the fish? To answer these questions, we examined individual fisher agent behaviour more closely.
A coastal fisher agent may not fish for several reasons: bad weather, no expectation of profits or wanting to spend time at home. Figure 6 (left) shows the reasons for coastal fisher agents not to go fishing over time. The main reason for not fishing is "spending time at home," varying between 10% and 20%. However, after approx. 15 years this need reduces and not expecting profits becomes the dominant reason for not going fishing and an increasing number of fishers stops fishing every year. If a coast fisher does not expect profits from a fishing trip, it will not go fishing. The expectations profits depend on the catch fisher agents had in the past. If a fisher had many instances of bad catches over time, the overall expectations of catch and profit will reduce to a point that the coastal fisher considers fishing as not profitable anymore and will stop fishing for the remainder of the year. As over time the fish stock decreases, for an increasing number of fishing trips, the visited spots are not good fishing spots anymore, see Figure  Experiment 2 (Section 4.1. 2) The effect of a fuel subsidy as a policy intervention was explored. It only affected the simulation runs with coastal fisher agents, wherein the subsidy triggered the fisher agents to fish more and consequently reduce the fish stock.
Surprisingly, we did not observe the same outcome in the simulation runs with the trawler fisher agents. We expected the subsidy to lower their costs and thus allow the trawler fishers to fish for longer periods, just like the coastal fisher agents. For the simulation runs with archipelago agents, we expected them to fish less, resulting in a higher fish stock, at the stable profit and satisfaction levels.
So, why did the fuel subsidy not influence the trawler and archipelago fishers? We explore this by exaggerating the fuel subsidy. Figure 8 shows the results of experiments to test the intuition that the subsidy reduces the cost and increases (expected) profits.
It shows that our expectation holds for the archipelago fishers. The introduction of a fuel subsidy lowers their costs, which means they need to catch less fish to be satisfied, resulting in higher fish stock levels. The subsidy thus helped to keep profits at the same level (compared to the pre-intervention situation), but with higher levels of fisher agent satisfaction and a higher fish stock.
For the simulation runs with trawler fisher agents, the introduction of a fuel subsidy had little effect because fish stock decline F I G U R E 6 Reflect the reasons for coastal fisher agents to not go fishing during the simulation experiments (left) and the ability of the coastal fishers to find the good spots for fishing (right). Each line represents the mean of 1,000 repetitions of the coastal fishery simulation Year Proportion of coastal fishers at good sites occurred so rapidly that the fuel subsidy only caused the trawler to go out more for a very short time. In the beginning, trawlers have high catches and high profits. Thus, they have no reason to expect lower catches and fish less. The subsidy then merely increases their profits (see Figure 8). The decline in profit resulting from lower catches, however, occurs so rapidly that the influence goes unnoticed. The subsidy amount cannot compensate for the bad catches because they are so low that costs are increasing too much. This demonstrates how the effect of a fuel subsidy depends on the situation of the fishers. The combination of the resource being (perceived to be) scarce, speed in which overexploitation takes place and the ability to actually catch more fish explains the limited influence of a fuel subsidy for the trawler fisher agents.

| D ISCUSS I ON AND CON CLUS I ON
In this paper, we present a way for integrating social scientific knowledge about individual and collective human behaviours into the modelling and management of fisheries. We demonstrate, using agent-based modelling, how such an integration is possible, and we thereby propose a scientific approach for reducing the uncertainty arising from human behaviour in fisheries. This approach lays the foundations for a next generation of social-ecological fishery models that account for aspects of human behavioural, social and ecological complexities that are purposive for a realistic assessment of a fishery sustainability problem.

| From acknowledging towards understanding
Scholars of fisheries science have taken an important first step in acknowledging the uncertainty they experience when making assumptions regarding fisher behaviour and social and ecological change in marine environments more generally. There exist fishery models that account for human behaviour more realistically (see Section 2.2). Yet, most of these models are based on expected utility maximization and focus on factors such as a fleet's ability to change strategies (van Putten, Gorton, et al., 2012) or on factors affecting fishing efforts, such as vessel or gear type. Only few models go beyond expected utility maximization and consider, for example, social norms or resistance to exit by relying on empirical data (Libre et al., 2015). We found no models that explicitly incorporate insights from social science theories of human behaviour. Altogether, this indicates a broad recognition of the need for more realistic representations and thus assessments across fields.

| Towards a next generation of social-ecological fishery models
We envision the next generation of fishery models to account for and enhance our understanding of the importance of the dynamics (and diversity) of human behaviour for the development and management of sustainable fisheries (Weber et al., 2019). These new models will be built on knowledge of human behavioural as well as biological and ecological complexities that is available from across the sciences. This knowledge will then be operationalized and contrasted against empirical findings or alternative models of similar fisheries.
Fisheries research and management benefits from a diversity of models that allow for studying and assessing fisheries (Nielsen et al., 2018;Weber et al., 2019). This diversity of model types is ultimately needed to advance the development of next-generation social-ecological fishery models. However, we see an important and specific role for ABMs in this development (Fulton et al., 2011;Lindkvist et al., 2020;van Putten, Kulmala, et al., 2012;Schill et al., 2019). Their main advantage is that ABMs allow for incorporating empirical or theoretical social science insight. It is their very nature to include both the microlevel (agents and their interactions) and the system or aggregate (macro) level (patterns) we aim to understand, such as overfishing. The necessity to specify agent characteristics and agent interactions in a social and ecological contexts over time allows for the incorporation of social science insights that may reside on microlevel, macrolevel or both level. Agent-based modelling is presently the only modelling approach able to reflect heterogeneous agents and their behavioural diversity over time (Conte & Paolucci, 2014

| New opportunities and roles of nextgeneration fishery models
These new models and the collaborative processes of their development can help address challenges of fisheries research and management in novel ways by (a) supporting a process of integrating knowledge across social and natural fishery sciences, (b) enabling the assessment of consequences of behavioural uncertainty and (c) serving as a means to identify underlying causal mechanisms that can provide entry points for governance.

Models as tools for integrating and contrasting knowledge about human behaviour
To create FIBE, we employed a joint development process, involving empirical scientists and modellers. The many conversations and iterations on the fishing styles and "bringing them alive" as agents exposed knowledge about context, style specifics and understanding beyond the written descriptions that reside between the lines of any conceptual/theoretical description of human behaviour. Moreover, it allowed us to tackle the inevitable conceptual gaps and logical inconsistencies encountered when formalizing theory (Sawyer, 2004 Apart from answering research questions with FIBE, we also explored other assumptions and theories to compare and contrast their effect on the outcomes. We tested, for instance, two ways for trawler and coastal fishers to be sensitive to the social information for decide where to fish: we formalized a "go where most others go" and "go where a (perceived) skilled fisher colleague is going." Although for this paper this exercise remained part of the sensitivity analysis, one might also implement competing explanations, observations or theories and explore their consequences.
In sum, model development and application can support a process in fishery management and research to bridge between social and ecological fishery science and can help improve the accessibility of behavioural insight for fishery research and management.
Modelling can serve as a central binding element and guide the process of integration between different fields, where the model becomes a common product, purpose, language and tool for mutual understanding.

Models for identifying causal mechanisms as entry points for governance
Fisheries are complex adaptive social-ecological systems, that are their intertwined dynamics of the social and the ecological system, their continuous interaction and influence on each other over time must not be neglected when seeking to understand why policies fail, overfishing occurs etc. We envision the next generation of fisheries models to not only help assess the implications of behavioural diversity and policy outcomes, but also to support analyses of underlying complex causal mechanisms that brought about certain outcomes (Biesbroek, Dupuis, & Wellstead, 2017;Schlüter et al., 2019).
Agent-based models such as FIBE lend themselves for opening the black box of human behaviour and investigating how certain outcomes or patterns came about. This is achieved through systematic experimentation with the model where certain processes are, for example, turned on/off, different behavioural models are tested or environmental or social contexts are changed. The analysis of model results then involves tracing the underlying processes leading to aggregate outcomes, such as overfishing. This process facilitates uncovering the different mechanisms that cause a certain outcome and to trace it back to the specifics of the different decision-making processes and their interaction within the fishery. With FIBE, we explored several why questions underlying the overall patterns (Section 4.2). For instance, we traced the reasons for not fishing to understand more why and at which moment the coastal fishers decide not go fishing, which could be either motivational (wish to be home was bigger than making profits), situation-based (bad weather) or experience-based (not expecting a profit).
The identified causal mechanisms may serve as anchor points for policies or interventions. They can help develop policy measures that are sensitive to the underlying processes and contextual aspects that give rise to fisher behaviour. Gaining a deeper understanding of why and how management measures such as a regulation are not effective is particularly relevant for managing complex systems such as fisheries. It may also point to measures that do not target individual fishers' behaviour, but rather influence social structures or other contextual conditions. Using models to unravel the causes of emergent outcomes is currently not a common use of models in management, but they can be of great importance when developing complexity-informed management and governance approaches.
For instance, our model has shown that the different types of behaviour generate different ecological, economic and psychological outcomes, as well as similar outcomes for different reasons. Both nuances are important when designing and implementing policy since the same policy can lead to different outcomes for the fishery and the fishers, or it may result in the same outcome but through different underlying mechanisms, with potential side-effects.

| Conclusion
We aimed to answer the calls to better account for human behavioural diversity in fisheries and provide support for others to do so. In this study, we took the next step by synthesizing knowledge from social (fishery) science and applying it to an exemplified fishery model. Social (fishery) science provides valuable insight into human behaviour and its underlying mechanisms and processes that are important in fishery contexts. While some fishery models are developing ways to include more realistic representations, most, however, lack in their approach to incorporate aspects of human behavioural diversity for understanding their implications for fishery outcomes.
This, would, however, ultimately help reduce and assess uncertainty as is required for advancing our scientific understanding of socialecological fishery systems. It would enable us to identify effective entry points for fishery management and thus improve overall management effectiveness with regard to sustainability of both the social and ecological systems. Our approach and findings highlight a promising avenue for reducing and assessing uncertainty based on human behavioural diversity by specifying, including and analysing its consequences through the use of agent-based modelling in fisheries (and beyond).

ACK N OWLED G EM ENTS
The research has received funding from the European Research  impressively improving the readability and accessibility of this paper and of each of the earlier manuscript versions; before (re)submitting, Gavin Masterson, who was our human filter for finding the fishery model papers for our review and Caroline Schill whose friendly and thorough review in the early days of the paper improved the manuscript tremendously; also, we thank our colleagues Gunnar Dressler, Marco Janssen, Andres Baeza, Karin Frank, Wander Jager, Birgit Müller, Nathan Rolling and Nina Schwarz, for stimulating discussions and creating a shared language in formalising behavioural theories in the early stages of this model during our "Human Decisions & Ecosystem Services" SESYNC project. Last but not least, we want to thank the anonymous reviewers for their thorough, thought-provoking and elaborate reviews on the earlier versions of this paper: they enabled us to improve this paper substantially.

DATA AVA I L A B I L I T Y S TAT E M E N T
The FIBE model is available on COMSES (https://www.comses. net/) allowing for the reproduction of all the findings of this study  https://www.comses.net/codeb ases/4d9cf ac5-7331-4a03-9d83-ab06c bedc1 43/relea ses/1.0.0/.
The simulation data, the r-scripts for analysis and summarized information from the literature review that support the findings are available from the corresponding author upon reasonable request.