Terrorist learning in context – the case of Al Qaeda in the Islamic Maghreb

ABSTRACT It is astonishing how many researchers adopt a counterterrorism agenda and suggest researching terrorist learning in order to shape security countermeasures. Posing different questions would lead to different answers. One such question would be, “What makes terrorist learning different?” Terrorist groups operate clandestinely, which means the environment in which they learn is different. This paper investigates the context in which Al Qaeda in the Islamic Maghreb (AQIM) has learned. Thus, a qualitative case study analysis of the influence of meso- and macro-level factors on AQIM’s tactical and strategic patterns between 1999 and 2013 will shed light on terrorist learning. Meso-level influences are conceptualised as cooperation and ultimate merging with Al Qaeda, and macro-level influences as government action. The result is puzzling: AQIM has learned tactically from Al Qaeda and strategically from counterterrorism. This is puzzling because scholars commonly question whether it is possible to learn under pressure. Nevertheless, AQIM’s learning has been more profound when faced with pressure than when cooperating voluntarily. The sustainable answer to the question of the political implication thereof is not how to boost counterterrorism measures but how to redefine them. If what is different about terrorist learning is above all the context, we need to question the context.


Introduction
In The Consequences of Counterterrorism, Crenshaw (2010) addresses the political consequences of counterterrorism and its negative impact on civil liberties. Counterterrorism often does not lead to what was intended and instead aggravates problems. When it comes to the subject of terrorist learning, it is astonishing how many researchers do not question but instead adopt counterterrorism agendas. Rasmussen and Hafez (2010, 3) suggest that research on barriers to innovation could help to shape security countermeasures; Kettle and Mumford (2017, 524) conclude that research on terrorist learning can help to expose and exploit weaknesses; and a Rand Report (Jackson et al. 2005, 59) suggests capitalising on the failures of terrorist groups by deceiving them about the impact of their operations, making "ineffective tactics appear more attractive or make successful techniques appear to fail".
Posing different questions would lead to different answers. Thus, Crenshaw argues that innovation must not be escalatory (Rasmussen and Hafez 2010, 8). Violence is often seen as the main feature of terrorism. However, terrorist learning need not imply radicalisationit can also lead to moderation. A further question that can lead to different answers asks what makes terrorist learning different. Carving out the distinction of terrorist learning would include highlighting its similarities and differences to the learning of other organisations. Interestingly, Kenney (2007, 144) writes about undercover operatives, suggesting similarities between secret service agents and members of terrorist groups when it comes to the element of secrecy. In fact, the main difference between terrorist groups and other organisations, such as social movements or companies, lies in their clandestine nature. This means that the environment in which they learn is different. Hafez (2018) has argued that "the defeat of violent jihadists usually follows from their own mistakes, not from the strategic talent of the powers that oppose them." However, these powers share responsibility for escalation. In fact, as this paper will demonstrate, Al Qaeda in the Islamic Maghreb (AQIM) learned strategically from the Algerian government. Counterterrorism led to learning. The sustainable answer to the question of the political implication of this is not how to boost counterterrorism measures but how to redefine them. If what is different about terrorist learning is above all the context, we need to question the context. This paper investigates the context in which AQIM has learned. The case study portrays a within-case comparison in order to contribute to a differentiated and more complex view of terrorist learning. Thus, a qualitative case study analysis of the influence of meso-and macro-level factors on AQIM's tactical and strategic patterns between 1999 and 2013 will shed light on terrorist learning. The newly established AQIM, resulting from the alignment of the Salafist Group for Combat and Preaching (GSPC) with the Al Qaeda core in 2007 and operating primarily in North and West Africa, provides an opportunity to examine terrorist group learning over time and may shed light on the underlying learning process of terrorist groups. Meso-level influences are conceptualised as cooperation and ultimate merging with Al Qaeda, and macro-level influences as government action. While the reaction to the merger with Al Qaeda provides an opportunity for cooperative, voluntary learning, the reaction to Algerian counterterrorism can be viewed as learning induced by pressure. The result is puzzling: AQIM has learned tactically from Al Qaeda and strategically from counterterrorism. This is puzzling because scholars commonly question whether it is possible to learn under pressure and they assume that organisations only act and react rather than learn when faced with pressure. Nevertheless, AQIM's learning has been more profound when they have been faced with pressure than when they have cooperated voluntarily.
It is not yet clear whether this result can be generalised and applied to other Al Qaeda affiliates. As Rollins (2011) stated, "Some affiliates receive money, training, and weapons; others look to the core leadership in Pakistan for strategic guidance, theological justification, and a larger narrative of global struggle." While this paper cannot claim generalisability, it aims to solve the puzzle that emanates from the gap between theoretical assumptions and empirical reality and thus to contribute to a more differentiated picture of terrorist learning. It thereby sheds light on how learning occurs under conditions of pressure and cooperation, and addresses what is more influential when it comes to the introduction of new tactics and strategies. It demonstrates that terrorist groups are able to learn and shows how conditions of pressure in the form of counterterrorism can be far more consequential than cooperation among terrorist groups. Since the status quocounterterrorismcontributes to terrorism, alternative modes of engaging with terrorist groups have to be designed. If context is the determining feature in terrorist learning, changing the context influences what is being learned. Practically, this opens the question of how the context in which terrorist groups find themselves can be changed to provoke lessons of de-radicalisation.
The paper begins with a review of the previous literature on terrorist group learning, followed by an analytical approach to terrorist group learning. The case study analysis focuses on the contextual factors impacting on AQIM's tactics and strategies. The concluding analysis elaborates on the weight of impact of the meso-and macro-level context.

Prior literature: terrorist group learning
In spite of the impact of learning on the longevity of terrorist movements, "surprisingly little work has been done thus far on the topic of tactical and/or technological innovation by terrorists themselves", as Dolnik has noted (2007, 10). However, especially since 9/11, the topic has gained increasing attention. In fact, the process in which terrorist groups learn and change has always been an element of the terrorism literature, much of which has been focused on whether terrorist movements are innovative or non-innovative (Jackson 2004). In 1993, Hoffman recognised that terrorist attacks are conservative in nature; in 1999, Merari concluded that terrorism had not changed much over the last century (1999,54); and in 2007, Dolnik himself argued that the scope of terrorist attacks is unchanging. The view that terrorism is conservative in nature also includes its strategies, as expressed for example by Bowyer (1990, 211), "Very few revolutionary organizations invest much time in strategic planning or organizational analysis. Any day-long analytical conference focuses deeper and longer on rebel strategy and tactics than do the rebels over a year." In contrast, Ranstorp and Normak (2015, 1) argued recently that terrorist groups are in fact highly creative and that some of them have the "propensity for remarkable innovation". Similarly, Crenshaw (Rasmussen and Hafez 2010, 8) questions whether terrorist innovations are indeed rare, and instead suggests that this misperception is due to several faulty assumptions such as that earlier methods do not qualify as innovations, that innovations by terrorists are necessarily escalatory, and that innovations are accompanied by their diffusion across other organisations.
The literature on the topic has advanced insofar as the debate over the innovative versus non-innovative nature of terrorist groups is increasingly being replaced with a discussion about the conditions for terrorist learning and the different kinds of terrorist learning. This discussion helps to reveal the complexity of the learning process. A major challenge with regard to the complexity of conditioning factors, levels of analyses and learning outcomes is to interrelate the variables and start theorizing. Not specifically writing about terrorism, Mumford and Hunter develop a multi-level perspective on creativity attempting to identify the relationships between variables and innovation (Mumford and Hunter 2005). With regard to terrorist learning Kettle and Mumford (2017) propose a new analytical framework for terrorist learning, and Singh (2017) develops a preliminary typology that maps pathways of learning and innovation by modern jihadist groups. While these are important advances when it comes to the subject of terrorist learning, the broader literature on learning has been characterised by Bennett and Howlett, who suggest that the concept of learning is still "overtheorized and underapplied" (1992,288). This is not the case with the study of terrorism and political violence, where some point to the predominance of case studies (Singh 2017). However, the conclusion to be drawn is not to abandon case studies but to use them instead for theoretical insights, which are best achieved through (within-case) comparisons: [A]s the data-driven study of terrorism and political violence slowly becomes more fine grained and comparative in nature, the focus has shifted to how (not just why) terrorism occurs. There is a creeping acknowledgment that innovation regularly occurs and can be categorized in a number of ways" (Gill et al. 2013, 129).
The question of how terrorists learn is essential to move beyond analyses that isolate motivations and outcomes. Addressing, for example, the question of how terrorist learning is different from the learning of other organisations problematises the context as well as the outcome of terrorist learning. There are several elements that are unique about terrorist learning. A major difference lies in the underground existence of terrorists. Thus, Crenshaw writes that the "commitment to violence as a primary method of action condemns terrorist organizations to…operate clandestinely. Terrorist organizations are predominantly underground conspiracies, and their activities are governed by the strictest rules of secrecy" (Crenshaw 2011, 69). Similarly, Horowitz writes that "[t]hey are different from states because they exist in a constant state of war" (2010b, 36). This specificity impacts upon the nature of learning processes that terrorist groups complete. The variables relevant for this paper, such as context conditioning in the form of counterterrorism and alliance with Al Qaeda, as well as learning outcome in the form of tactics and strategies will contribute to our understanding of how terrorist learning is different.

The context of terrorist learning
While Mumford and Hunter paint a complex picture of organisational innovation, they also pledge for differentiation and ask, "Which forms of change are more likely to stimulate innovation on the part of organizations?" (2005, 50) In a similar vein, Crenshaw poses the question: "Is it possible to distinguish sources of innovation according to type?" (Rasmussen and Hafez 2010, 42) In the literature, the context of terrorist learning is roughly differentiated as being either internal or external.
With regard to the internal context, multiple authors draw an analogy between business organisations and terrorist groups (Takeyh and Gvosdev 2002, 97-108), applying vocabulary and concepts from the business literature to the study of the evolution of terrorism. For example, researchers at the RAND Corporation have advanced understanding of terrorist learning by applying models of organisational learning to the subject (Jackson 2004;Jackson et al. 2005). The lens of organisational learning can contribute to the study of terrorist learning through its focus on a collective process. Knowledge becomes organisational and does not depend on particular individuals to be utilised and implemented (Jackson 2004). In Aptitude for Destruction, Jackson et al. (2005, xi) specify the component processes of organisational learning as acquisition, interpretation, storage and distribution of knowledge. Trujillo and Jackson (2006) further specify the determinants of organisational learning as including structure, culture, knowledge resources and environment. These component processes and determinants predominantly represent internal dynamics within terrorist organisations.
Although this approach is useful, learning does not occur in a vacuum, and there are external contextual influences that impact on terrorist group learning and adaptation as well, as I will show. The notion of social learning provides an applicable conceptual tool for contextual analysis. As Bandura states, "Developments in learning theory shifted the focus of causal analysis from hypothesized inner determinants to detailed examination of external influences on responsiveness." (1977,2) According to Bandura, learning occurs through experience or through the observation of others' behaviour. While social learning has frequently been applied to the study of individual behaviour, social movement theory provides an analysis of collective behaviour that equally focuses on external influences and thus context. Among the authors who have applied social movement theory to the study of political violence, Della Porta (2006) argues that understanding radical political movements requires analysing the micro, meso and macro level. As Gunning notes, "[t]he overall result [of applying social movement theory to the study of terrorism] is that terrorist violence is de-exceptionalized, taken out of its sterile box of sui generis phenomena, and returned to its living context" (2009,162).
The meso level thereby refers to relationships with other terrorist movements and is decisive for terrorist learning. For example, as Horowitz (2010b, 166) has observed, [S]ometimes desire is not enough to adopt an innovation. Even though Al Qaeda had money, committed members, and weapons, it sent its members to Hezbollah, a suicide attack innovator, to pick up thetacit knowledge necessary to conduct its own operations.
Several scholars connect cooperation among terrorist groups with the diffusion and escalation of violence. For example, they have reported evidence of the global contagion of terrorism as well as the increased lethality of terrorist attacks (Sandler 2014, 257-271;Midlarsky, Crenshaw, and Yoshida 1980, 262-298). Horowitz and Potter (2014, 199-225) argue that lethality is affected by meso-level factors such as terrorist intergroup cooperation. Rather than operating in isolation, as they observe, terrorist groups forge alliances that increase their ability to conduct attacks. Similarly, Pedahzur and Perliger (2006, 1987 argue that transcending social networks impacts on the use of suicide terrorism. Other scholars correlate competition among groups with an increase in violence. Thus, McCormick concludes that "the currency of…competition is violence" (2003,488). However, as Dolnik notes, "[T]he motivations of terrorist organizations are much more diverse and much less linear. The success of terrorist operations cannot simply be measured by the number of casualties inflicted" (2007, 10). Alternatively, Bloom (2004, 61-88) reasons that terrorist organisations compete with each other for public support and market share. She connects the use of Palestinian suicide bombings to the competition of Palestinian terrorist groups for recruitment. Further scholars suggest that the primary motives for terrorist group alliances are driven by survival (Karmon 2005;Blomberg, Gaibulloev, and Sandler 2011, 441-463).
As a further external contextual factor, macro-level forces equally affect terrorist group learning. Some scholars have correlated counterterrorism with terrorist innovation, especially tactical innovation. For example, Dolnik (2015) writes that "greater frequency of attacks is likely to have a profound impact on both the desire as well as the capability of terrorists to innovate". Similarly, Ackerman (2015) writes about the "commonly described 'offense-defense' co-evolutionary dynamic" between terrorist groups and states. This dynamic is part of broader dynamics that "create almost Darwinian pressures for a terrorist organizationit must adapt and evolve if it wants to enjoy any type of longevity." While the reaction to the state is often framed as tactical in nature, some scholars point to more profound changes: "But if the environment shifts considerably, a group needs to make more dramatic changes in order to survive. In this case, a group may have to pursue 'discontinuous change'learning aimed at entirely new activities or making large shifts in the way the group plans and acts." (Jackson et al. 2005, 15) Scholars of Social Movement Theory have outlined co-evolutionary dynamics of a more profound nature. Tarrow (1994, 153) argues that actions by the government stimulate innovation in social movements and their strategies. With regard to terrorist organisations, Waldmann (2005) describes the competition between terrorist groups and the state as motivated by terrorists' desire for public attention and provocation. While the co-evolutionary dynamic between states and terrorist groups can be seen as a driver of innovation, a hostile environment does limit learning capability.

Outcomes of learning
Putting terrorist learning in context means understanding which contextual conditions trigger kinds of learning. The literature to date has observed that terrorist learning is mainly incremental in nature rather than revolutionary. This incremental character is correlated with the environment in which terrorist groups find themselves: "[I]nnovation itself is often incremental and driven by the need to overcome security constraints" (Gill et al. 2013, 129). Not only are terrorist innovations built upon old ways, they are additionally frequently reactive in nature: "With regard to tactical innovations, most analysts agree that terrorists respond to what the government is doing…It might also be imitation of what competitors are doing. Reactivity would be the critical distinction between strategic and tactical innovation." (Rasmussen and Hafez 2010, 43) Crenshaw's distinction between tactical, strategic and organisational innovations is thereby often referred to as a helpful frame in the literature. She defines the three kinds of innovation as follows: Tactical innovation usually involves inventing or adopting new techniques or technologies to achieve unchanging objectives. Strategic innovation entails formulating new objectives, which necessitate the adoption of new operations, targets, or technologies to advance those objectives. Organizational innovation involves new ways of structuring the terrorist group or inventive methods of drawing recruits. (Rasmussen and Hafez 2010, 1) Whereas this distinction is often taken up by other scholars, much of the work on terrorist learning has employed an instrumental understanding of the evolution of terrorists by focusing on tactical learning. Conclusively, various authors are focused on the increasing lethality of terrorist attacks as a result of terrorist innovations. So-called "new terrorism" or "religious terrorism" is thereby frequently correlated with more lethal and catastrophic attacks (Laqueur 2000;Crenshaw 2000, 405-420;Hoffman 1997, 1-15). After 9/11, the adoption of suicide terrorism thus gained widespread attention in the literature. 1 However, well before the attacks on the twin towers, terrorism had begun to transform, revealing two major trends: the shift from harder to softer targets in the wake of increased counterterrorism measures, and the internationalisation of terrorist attacks (Hoffman 2006, 63).
Looking at tactical evolutions in isolation from the strategic developments of terrorist groups can lead to overlooking these broader contexts. In addition to tactical learning, this paper is designed to include an analysis of strategic learning. Therewith, it interrelates conditioning variables (meso and macro level) and learning outcomes (tactical and strategic), thus addressing the gap between the theoretical and empirical realm.

The puzzle
Crenshawwhose tripartite differentiation of innovation as tactical, strategic and operational is useful for framing terrorist learningalso notes that "it is not easy to distinguish between tactical and strategic innovation and between innovation and adaptation" (Rasmussen and Hafez 2010, 38). She sees the element of reactivity as the critical distinguishing feature between tactics and strategies. Albeit scholars dispute that terrorist learning is simply an automated response to environmental factors, terrorist groups are confronted with forces that might limit their learning capability beyond reactivity. Kenney for example describes how operating underground can result in "'burn syndrome,' the pervasive fear that other people know what they are doing" (2007,145). The impossibility to exit an underground organisation except when willing to risk one's life creates a complex situation in which learning beyond adaptation and automated reaction seems difficult to imagine. As has been pointed out by scholars, pressure only leads to a change in routine behaviour, and what is learnt is not internalised (Schwabe and Wolf 2009). In fact, the environment in which terrorist groups find themselves poses a special challenge to learning. In Forrest's Teaching Terror, Trujillo and Jackson (2006, 62) speak of environmental uncertainty that affects the learning capacity of terrorist groups. In a Rand Report, Mumford and Hunter (2005, 58) refer to causal ambiguity, which describes how strategic planners have difficulty determining the impact of their choices because of the complexity of the environment. This, they argue, is especially the case for groups operating in secrecy and under high levels of threat. The ambiguity of causal connections leaves decision makers with the possibility of accidental learning. According to Holyoak and Thagard (1995, 192), sources of learning and problem solving are often noticed by decision makers "through serendipity or accident". They further specify that innovative solutions, "mental leaps", arise when one is faced with a new situation.
Cognitive flaws can explain why causal ambiguity can be aggravated by misperceptions. Hafez finds that jihadists in particular are incapable of learning due to cognitive flaws: Several cognitive and organizational mechanisms can help explain why ideological extremists cannot compromise with rival groups…ideologically extreme groups are likely to associate with other extremists, leading to an ideological encapsulation that shuts out the countervailing voices necessary to learning and adapting" (2018, 88).
The connection between ideology and learning has been addressed by some scholars. Drake asserted that the ideology of a group is key to its interpretation of the world and its definition of its enemy (1998, 54); Dolnik has argued that a group will be provoked to innovate when it changes its worldviews (2007,; and Moghadam (2008) has found that Salafism was the driver behind the use of suicide attacks by Al Qaeda. In fact, innovation is correlated with groups that are highly risk tolerant. Interestingly, groups with very conservative ideologies can use very modern tactics (Dolnik 2015, 76-80).
What is puzzling is that in the case study depicted in this paper, an Islamist terrorist group learned strategically through its interaction with the state and learned tactically from allying with Al Qaeda. Whereas the learning capability of organisations is sometimes questioned, hostile environments complicate learning of terrorist organizations further. In such cases, scholars highlight the difficulty of strategic learning in hostile environments instead of the more automated response in form of tactical learning. And yet, AQIM learned strategically from counterterrorism. What is more, AQIMan Islamist group belonging to the category of jihadists that Hafez finds incapable of learninglearned profoundly. This paper tries to solve this puzzle.
Analysis: the case of al qaeda in the islamic maghreb (AQIM) The case of AQIM provides an example of terrorist learning that can be analysed to examine the impact of both meso-and macro-level influences on changes in terrorist group tactics and strategies over time. In the following section, a brief overview of AQIM's historical background is provided and AQIM's tactical evolution and strategic shifts are subsequently analysed.

Background of GSPC and AQIM
The development of Islamism in Algeria dates back to the 1980s when many Algerians decided to join the jihad against the Soviets in Afghanistan. Their return from the Afghani battlefield coincided with the rise of Islamism in Algeria due to repression by the Algerian government. Soon the GIA was formed, gaining notoriety for its terrorism against the Algerian population: "Within the broader context of the civil war which engrossed Algeria in the 1990s, GIA helped contribute to over 100,000 civilian deaths during that time" (Colvin 2008, 4). According to the GIA, apostasy was not limited to the Algerian government but extended to society as a whole. Members within the GIA who did not agree to the indiscriminate targeting of civilians departed to create the Salafist Group for Preaching and Combat (GSPC) in 1998. The GSPC declared an end to the attacks on civilians. Its fight against the state pushed the Algerian government to embark on a course of repression and infiltration of the GSPC. The Algerian government's counterterrorism programme was viewed as largely successful (Larémont 2011, 242-268). Repression was only exacerbated with 9/11. The war on terror augmented the pressure on the GSPC: As early as 2001, there existed a widespread view among European security services that the GSPC was on the verge of collapse. Similarly, Algeria's security establishment from late 2001 tried to reframe the struggle against Algeria's armed groups as part of the international 'war on terrorism'and claimed that the disappearance of the Algerian terrorist groups was merely a question of time. (Steinberg and Werenfels 2007, 407-413) Repression weakened the group, resulting in dissension and frequent leadership changes. In 2004 the GSPC started reaching out to Abu Musab al-Zarqawi, Al Qaeda's Iraqi leader. Via the contribution of Algerian fighters to the Iraqi struggle, the Al Qaedaminded factions within the GSPC sought to merge with the global jihad. According to several scholars, the GSPC sought to ally with Al Qaeda as early as 2003. However, Ayman Al-Zawahiri 2 approved the merger only in 2006. In January 2007 the GSPC renamed itself to Al Qaeda in the Islamic Maghreb (AQIM). According to Steinberg and Werenfels (2007), the GSPC's weakness prompted it to seek the alliance with Al Qaeda. The year of the merger became the most violent year since the Algerian civil war.
The implications of the alliance with Al Qaeda deserve attention. An analysis thereof can shed light on how the merger impacted upon tactics and strategies of the North African Islamists. Furthermore, this impact can be compared to AQIM's reactions to government repression. The case of AQIM is illustrative of the impact of the meso and macro level on terrorist learning.
Tactics: the introduction of suicide bombings AQIM has introduced new tactics since its merger with Al Qaeda in 2007. Most noticeably, the group has adopted suicide bombings. For example, in 2007 forty-one people were killed by a double suicide bombing in Algiers and this attack was not the only suicide bombing that year. Following an estimate by Jean-Luc Marret, ten suicide attacks occurred in 2007. Marret also notes the increase in the number of fatalities (2008,(541)(542)(543)(544)(545)(546)(547)(548)(549)(550)(551)(552). The suicide bombings perpetrated have led to a higher number of victims. The use of suicide bombings is a new phenomenon in North Africa: It ought to be noted that suicide attacks were very rare in North Africa, even during the height of the Algerian civil war which left hundreds of thousands dead or wounded, until more recent years when 'legitimized by ideologues close to al-Qaeda, it became increasingly commonplace'. (Pham 2011, 240-254) Other tactical developments equally hint to Al Qaeda's influence. The use of internet videos is a practice adopted via jihadi networks (Marret 2008, 541-552). Furthermore, Al Qaeda has trained members of its local branch in North Africa in the use of explosives. AQIM has used improvised explosive devices, similar to tactics employed by Al Qaeda. Another innovation is the introduction of synchronised attacks. AQIM's first attack in 2007 was perpetrated via the simultaneous detonation of seven car bombs (Gray and Stockham 2008, 091-097). Arguably, AQIM's new tactics are a hallmark of operations of the global jihad. The introduction of suicide bombings marks a watershed in the history of GSPC and AQIM (Table 1) 3 .
When observing the evolution of the use of suicide attacks by AQIM, it becomes apparent that the sudden leap in suicide attacks that resulted from the merger with Al Qaeda fades out after 2008. This is telling as to the short-term versus long-term character of this tactical innovation. Clearly, course corrections concerning tactics are easier than course corrections concerning strategies. The momentum of something new is, moreover, hard to sustain since exponential progress might be required to keep that momentum going. Furthermore, the fact that AQIM halted the attacks could speak to a gap between external influence and internal convictions. If the alliance with Al Qaeda was a marriage of convenience, profound internal changes cannot be taken for granted. Instead, the empirical facts speak to pragmatism. Although a name change is profound, it does not necessarily reflect a simultaneous change in preferences and interests.
Since its merger with Al Qaeda, the Algerian Islamists have introduced new tactics. This tactical learning reveals the influence of meso-level factors. AQIM has been able to learn tactically through its association with Al Qaeda. What, in turn, explains AQIM's changing strategies? In order to answer this, the next two sections examine strategic shifts in involvement in guerrilla warfare versus terrorism, and the emphasis on the near versus the far enemy.

Shifting strategies: guerrilla warfare versus terrorism
The GSPC was formed as a reaction to strategic mistakes of the GIA, its predecessor. The GIA's doctrine of apostasy of the entire society alienated the population. Fear and indoctrination through punishment of non-conforming civilians were commonplace (Gray and Stockham 2008, 091-097). The failure of the GIA has to be viewed in the context of Algeria's counterterrorism efforts. A military offensive combined with civilian assets contributed to turning "the tide against the rebels" (Echeverría Jesús 2004, 354). In the second half of the 1990s, Algeria based its anti-terrorist strategy on three pillars: a military offensive, a political offensive and a propaganda war (Jesús, Carlos 2004, 354). As a result, the GIA only intensified its slaughtering of civilians. The loss of support by the population led to division within the Islamist movement, which the government was able to capitalise upon (Jesús, Carlos 2004, 354). Moreover, the government introduced a division of labour: While the president's main task was to campaign for the elections, the army and security services would intensify their operations to ensure maximum security for them. Moderate Islamist candidates steered a careful path between their Islamist identity and respect for the constitution and electoral laws, which forbid parties based on religion and the use of mosques for political activities. (Jesús, Carlos 2004, 354)   However, the strategic mistake by the GIA did not lead to the disappearance of radical Islamism in Algeria. Hassan Hattab formed the GSPC in 1998, basing its doctrine on the rejection of attacks against civilians. Consequently, the GSPC confined its attacks to security forces and state institutions. The strategic shift from terrorism to guerrilla warfare was a reaction to the weakness of the GIA and the success of the government. A comparison of the impact of counterterrorism with the impact of the merger with Al Qaeda is revealing; it opens the question of whether the shift to guerrilla warfare has been reversed and whether attacks against civilians resumed after the group allied with Al Qaeda. Since the merger, civilians have been attacked and according to Marret (2008, 541-552), the GSPC has been unable to continue guerrilla warfare due to difficulties and losses in the country and especially to the efficiency of counterterrorist activities. In fact, AQIM's attacks against civilians were preceded by similar attacks by the GSPC. Furthermore, AQIM continued to distance itself from the killing of civilians: "We are not insane to target our Muslim Brothers" was a statement made by Abdelmalek Droukdal, one of AQIM's leaders in a New York Times interview. He was reacting to a question about AQIM's attack on the regional UN headquarters and the Algerian Constitutional Court in Algiers on 11 April 2007. The attack in Algiers had resulted in forty-two casualties and 132 wounded civilians. Droukdal further stated, "Mistakes may occur in the war but they are unintentional" (Colvin 2008, 4). Judging by these statements, AQIM did not shift its strategy due to the merger. The strategic shift to attacking the state can be seen in the context of government pressure, and the resumed attacks on civilians preceded the merger with Al Qaeda. Nevertheless, since the merger with Al Qaeda, foreign civilians have been attacked. This raises the issue of a further strategic decision that could indicate strategic learning as a result of interaction at the meso-level: attacking the near or the far enemy. First, international theatres of operation are not a new phenomenon. The involvement in the Iraqi war is an example of international engagement that began before the rebranding of the GSPC as AQIM. The GIA targeted foreign civilians prior to merging with Al Qaeda. Furthermore, in 1994, the group hijacked an Air France flight and in 1995 it attacked civilians with bombs in Paris (Gray and Stockham 2008, 091-097). The kidnapping of international tourists is not a new technique either. As an example, the GSPC abducted thirty-two German, Austrian and Dutch tourists in 2003. In fact, out of sixty-three Westerners kidnapped between 2003 and 2011, thirty-two were seized in 2003 (Larémont 2011, 242-268). However, according to Pham, [T]he focus on Western prisoners and extraction of spectacular ransom payments and concessions can certainly be said to represent a qualitative jump that was facilitated by both the 'rebranding' of the organization as an affiliate of al-Qaeda in 2006 and the ideological influence which accompanied it. (2011,246) Since AQIM has perpetrated few attacks outside of Algeria, Filiu (2010, 12-13) argues that AQIM began attacking global targets locally. Targeting the far enemy within the local environment might suggest a strategic shift. However, the kidnapping of Western hostages precedes the merger between the GSPC and Al Qaeda. Furthermore, a key strategic difference between AQIM and Al Qaeda can be observed. Al Qaeda asks for political demands when taking hostages. AQIM conducts kidnappings with the goal of collecting ransom. As stated by Filiu, Al Qaeda only successfully pressured AQIM into killing its hostages once, in the case of a British hostage in 2009, and even then AQIM did not "give the killing Zarqawi-like publicity" (13). The kidnappings by AQIM can be seen as a means to an end rather than as speaking to a strategic shift.
Second, attacks against American interests are rare. According to Marret (2008, 541-552), the attack on a bus carrying employees of a subcontractor of Halliburton, a company that is active in Iraq, marked a turning point. Furthermore, Steinberg and Werenfels note AQIM's change of rhetoric. The U.S. and France are far enemies that are increasingly threatened in public declarations (Steinberg and Werenfels 2007, 407-413). However, the change has been merely rhetorical. Rollins (2011) observes an increase in the anti-Western rhetoric of AQIM as well but equally observes that its attacks remain focused on Algeria and the Sahel. The attacks against the far enemy represent a minor portion of AQIM's terrorist activities (Torres Soriano 2011, 279-298).
Clearly, the group is focused on attacking the near enemy: Algeria. The regional enlargement of AQIM's theatre to include parts of Mali, Mauritania, Niger and Chad are likely a result of counterterrorism operations. The group has been increasingly pushed out of its original area of operation (Larémont 2011, 242-268) and is increasingly involved in criminal activities. Thus, "many governments and analysts argu[e] that its radical Islamist rhetoric merely serves as a cover for the group's profitable criminal activities" (Boeke 2016). As Boeke (2016) points out, the policy implications are clear to Western decision makers, namely that an increased international effort must be pursued to disrupt illegal trafficking. However, AQIM's weakening and hence operations in Mali and exploration of criminal ties have been a result of government pressure. This government pressure is not only exercised by the Algerian government but includes international pressures such as the French military intervention in Mali in 2013 and U.S. support for Algeria. Steinberg and Werenfels (2007, 407-413) argue that the weakening of the GSPC has also been a result of American support for Algeria's security forces since 9/11. Similarly, Marret states that "[t]he efficiency of Algerian and European counterterrorist services contributed to push the GSPC away to the Sahara, Darfur or Iraq" (2008,545). Moreover, the far enemy of Al Qaeda translates into attacks on American targets and interests. In its declarations, AQIM most frequently cites France as an enemy associated with Algerian interests. Finally, in his analysis of the evolution of the discourse of AQIM, Manuel Soriano comes to the conclusion that the rhetoric of the group has not become more global and that global content already existed before the merger with Al Qaeda. He argues that the messages regarding Muslims abroad have evolved according to the events in these countries rather than reflecting acceptance of Al Qaeda's agenda (Torres Soriano 2011, 279-298).
Third, the (limited) internationalisation of the GSPC can be seen as the cause of the merger with Al Qaeda and not its outcome. Via Iraq, the GSPC reached out to Al Qaeda. Larémont argues that Droukdel's "view of jihad was compatible with those of Osama bin Laden and Ayman al-Zawahiri. Because of their convergence of interests, Droukdel decided to merge his group to Al Qaeda Central" (Larémont 2011, 244). According to several sources, approaching Al Qaeda was caused by the weakness of the GSPC: "Islamist failures in the Maghreb have created a loss of dynamism in the movement. Given the ideological and organizational crisis, the Maghrebi jihadist movement was in need of rejuvenation." (Celso 2008, 91) Arguably, the search for an international agenda began when the GSPC "was pushed onto the defensive" (Steinberg and Werenfels 2007, 409). American support for Algerian counterterrorism has only contributed to the adoption of an international agenda, according to . The framing of the war on terror as a crusade by President Bush might have been an additional push for the GSPC to perceive itself as a global terror group: "The comments [by Bush] were a catalyst for a movement that was lacking at the time in cohesiveness or a unifying ideology" (Colvin 2008, 12). According to Steinberg, the GSPC was looking to join international groups because it was weakened. It merged with Al Qaeda in order to win support against the near enemy but it did not want to shift strategies (Steinberg 2005). Filiu similarly states that "the continuity from 2003 until today between the GSPC and AQIM has been striking. Despite the inflamed global rhetoric and the Internet echo chamber, the main difference lies in the generalization of the suicide attacks" (2009,224).
There are several arguments that indicate meso-level contextual factors did not result in a strategic shift for AQIM. First, attacks against Westerners preceded the merger with Al Qaeda. Second, AQIM's focus on the far enemy remains marginal and trumped by regional interests. Third, AQIM's limited international focus can be seen as a result of government pressure that caused the merger with Al Qaeda rather than resulting from it. Macro-level contextual influences such as government pressure have been more decisive in causing a change in strategy.

Concluding analysis
Is it possible to learn under pressure? Crenshaw sees the element of reactivity as the critical distinguishing feature between tactics and strategies (Rasmussen and Hafez 2010, 43). The complexity of a threatening environment complicates learning, especially strategic learning. Cognitive flaws and misperception due to causal ambiguity are moreover aggravated when it comes to ideologically extreme groups, Hafez (2018) finds. These groups live in a filter bubble, so to speak. The case study depicted in this paper is an Islamist terrorist group that learned strategically through its interaction with the state and learned tactically from allying with Al Qaeda. That the group demonstrated the capability of strategic learning in a hostile environment is puzzling. Researchers from different disciplines have observed that pressure only leads to a change in routine behaviour and that what is learnt is not internalised (Schwabe and Wolf 2009). The new behaviour is, moreover, frequently unlearned and the changes are therefore shortterm rather than long-term. This leads to the paradox that "the need to learn from mistakes is highest in times of political scrutiny; on the other hand, under such conditions learning is presumed to be most difficult" (Dekker and Hansén 2004, 211).
The case of AQIM contradicts the assumption that terrorist organisations are only able to act and react rather than learn. Whereas being forced to change can lead to blaming and scapegoating rather than internal reflection, it can also provide a window of opportunity for leaders who have ideas for change and need an argument in order to sell the change to others (Dekker and Dan 2004). In this way, pressure usually does not lead to new insights but provides an argument for change based on already existing insights.
With regard to terrorist organisations, the rules of their decision making are also defined by the conflict. When the conditions for learning on a meta-level are lacking, terrorist organisations still learn within the parameters of the conflict and change their tactics and strategies. Theoretically, pressure should more easily be connected to a change in tactics, whereas strategies should be more robust, unless the idea for strategic shifts was already there in which case pressure is a welcome window of opportunity. Furthermore, organisations can follow a trial-and-error approach: "Learners become experimenters who take initiatives to explore their habitats…they begin to understand their environments and become more able to manipulate and change their situations" (Hedberg 1981, 5). In contrast, "[l]earners who only learn to respond and who imitate others behavior have little power over their environments" (Hedberg 1981, 5). The ability to affect the environment, for example supporting communities, is at the heart of the conflict between states and terrorist groups. This paper has addressed the question of whether pressure or cooperation is more influential when it comes to the introduction of new tactics and strategies. In order to answer the question, terrorist group learning was investigated by examining changes in AQIM's tactics and strategies over time. Whereas AQIM's conflict with the state is characterised by pressure and threat, the relationship with Al Qaeda is based on more cooperative terms. Throughout its course of evolution from the time of the GIA to the founding of the GSPC and its merger with Al Qaeda, the Algerian Islamist group(s) has learned tactically and strategically. In addition to minor tactical innovations, it has established a new influential tactical routine that has changed the operational conduct of the conflict: suicide bombings. The use of this tactic has been a new development in North Africa. However, the GSPC/AQIM also learned on the strategic level. The GSPC was founded by Hattab to correct past mistakes. The leader rejected the killing of civilians and completed a strategic shift from terrorism to guerrilla warfare. This development clearly marks a shift that represents a new and consequential routine. The (limited) shift to attacking the far enemy within the local environment constitutes a further strategic choice with wide-ranging implications, for example for American and European threat perceptions.
Terrorist learning has been framed in this paper as occurring in a particular context, with an emphasis on the impact of government pressure (macro level) versus cooperation with other terrorist groups (meso level). The goal was to discern how the meso and macro level impacted upon the tactical and strategic learning of the group. The GSPC/AQIM has learned tactically from Al Qaeda. The introduction of suicide bombings can be seen in the context of the merger. In 2007 alone, the year of the rebranding of the GSPC to Al Qaeda in the Islamic Maghreb, Marret reports that AQIM perpetrated ten suicide bombings and the GTD reports twenty-five attributed suicide bombings to AQIM between 2007 and 2013. Furthermore, suicide attacks are a trademark of Al Qaeda. Further tactical innovations such as synchronised attacks and the use of internet videos for propaganda purposes equally reveal the influence of Al Qaeda on AQIM's tactical evolution.
Interestingly, when depicting AQIM's suicide attacks in a graph, a momentum can be observed that fades out again after 2008. The tactical change has not been a long-term innovation. Several possible explanations come to mind: the suicide attacks were perhaps less successful than assumed; they resulted in unintended consequences; or there was a mismatch between internal understandings and new tactics. What this case reveals is that cooperation is also limited in its impact on internal learning processes. When reactions to external influences are not accompanied by internal shifts, they are rarely long-lasting.
AQIM has not learned strategically from Al Qaeda. Instead, its shifts in strategies were induced through pressure by the government. Thus, the GSPC changed strategies because the GIA had been decisively weakened by the Algerian government. The rejection of the doctrine of the apostate society was a reaction to a strategic mistake by the GIA that was consequential for the GIA's support in the population and its position vis-à-vis the state. Hassan Hattab founded the GSPC to correct this mistake.
When learning under pressure, what is learned is frequently a by-productthe main goal is to stop the pressure. Sometimes pressure causes the pressured to unlearn behaviour. However, once the pressure is lifted, the unlearned by-product can reoccur. The question of long-term versus short-term change in behaviour is relevant for a predictive analysis of terrorism, for example for the consideration of whether the GSPC has only unlearned attacking civilians as a by-product of pressure. A related question concerns what is learned to replace the unlearning. In fact, unlearning is necessary for learning. According to March and Olson, there are times when organisations should treat their memories as enemies (March and Olson 1976). One way of learning in organisations is simply to fire high-ranking people. With Hassan Hattab a new leader came along who brought with him the idea for change. The pressure by the Algerian government assumingly provided a welcome window of opportunity, an argument to realise the change. It has to be considered that mergers with other terrorist groups can also lead to an influx of new personnel, opening doors for new kinds of learning. In the case of AQIM, the new learning from the merger with Al Qaeda came in the form of tactical learning. It can be assumed that strategic learning takes place predominantly on the part of the leaders.
The second strategic choice, attacking the far enemy within the local environment, equally resulted from government pressure. Pressure by the government can hence serve both as an argument for de-radicalisation (renouncing the killing of civilians) and as an argument for radicalisation (attacking the far enemy). In fact, the action-reaction circles between terrorism and counterterrorism have frequently been elaborated upon, for example with the argument of terrorism as a provocation (Waldmann 2005). Terrorist groups hence experiment with their environment and provoke the state into repression in order to win the people for their cause. Such trial-and-error experiments go beyond simple adaptation: "Learners become experimenters" (Hedberg 1981, 4).
In the case of AQIM, increased counterterrorism efforts after 9/11, including American and European support for the Algerian government's counterterrorism efforts, weakened the GSPC, which attempted to merge with Al Qaeda as a consequence. As a result, the GSPC tried to internationalise its efforts, mostly in rhetoric. The alliance with global jihad was a result of changed strategies, not a motivation for strategic learning. As Celso has concluded: "Al Qaeda in the Maghreb represents a marriage of convenience" (2008, 95).
While the meso-level impacted upon tactics, the macro-level impacted upon strategies. It can therefore be concluded that government pressure has been more consequential than the merger with Al Qaeda. While AQIM learned tactically and changed its methods as a result of the alliance with Al Qaeda, it changed its strategies as a result of force. While strategies are more robust, AQIM has exhibited flexibility and ease regarding tactical innovations.
Several conclusions can be drawn from these observations. One of these concerns the international threat that AQIM poses. While some scholars perceive AQIM as a serious threat to the U.S. (Isaacs 2009) as well as Europe, others see its influence confined to local commitments in North Africa. The threat posed by the group, as argued in this paper, has not predominantly increased through the cooperation with global jihad but instead is related to government pressure. In fact, instead of asking which threat AQIM poses, it might be useful to ask which threat counterterrorism poses. Understanding the co-evolutionary dynamic between states and terrorist groups could help to think about a learning context that facilitates de-radicalisation. However, that would have to imply acknowledging the responsibility of states in the spirals of violence and grasping that it is precisely the context terrorist groups find themselves in that explains their behaviour.
The external context of terrorist groups is thereby easier to study and potentially easier to influence by decision makers than the internal environment. However, the internal and external environment are closely related. The case of the GSPC/AQIM reveals a greater flexibility when it comes to tactical learning. The opportunity to acquire knowledge about new tactics, such as suicide attacks as a result of the merger with Al Qaeda, has been seized by the group. In terms of strategies, it exhibited more rigidity and only changed when forced to by macro-level influences. A possible explanation is that internal resistance to change can only be overcome when faced with substantial pressure. Leaders who want to implement changes need arguments for course corrections. Thus, the question arises: What happens inside such organisations and how do we link internal dynamics to external influences? Focusing only on the internal dynamics at the micro-level or on the external context leads to different questions, whereas understanding their relationship can help to address the gaps that remain when answering these questions separately. Interrelating the levels would shed light, for example, on the internal processing of information that is influenced by outside pressure, or on how the leadership communicates pressure from the outside as well as cooperation with other groups internally. Although these questions go beyond the scope of this paper, an assumption underlying this research is that thinking about the different levels together is beneficial indeed.
The internal dynamics of AQIM are surely influenced by its ideology. That AQIM can be characterised by what Hafez (2018) describes as ideological encapsulation poses the puzzling question of how strategic learning became possible nevertheless. As Hafez (2018) asserts, ideologically extreme groups frame the world in terms of us versus them. What is interesting is that frequently groups learn the most from their enemies 4 and that learning from enemies can lead to profound insights into reality and one's weaknesses. Furthermore, it could be possible that leaders can sell their ideas for change internally using the argument of external pressure, and that this is especially possible for ideologically extreme groups in which the portrayal of the enemy is so important. In fact, these arguments contradict Hafez's view that ideologically extreme groups are incapable of learning, and underline that very conservative groups are not only able to use very modern tactics but also novel strategies. Moreover, novel strategies can be developed even when facing a hostile environment. If counterterrorism efforts yield the arguments for strategic learning, government action has to be rethought. While the impact of allying with global jihad was limited to tactical innovations in this case, the strategic reactions to counterterrorism were of a far more consequential nature.