The advances in the field of argumentation theory and the research in the area of education are increasingly becoming interconnected (Mirza Muller and Perret-Clermont 2009). On the one hand, education is the natural and traditional application of argumentation studies. Since Plato’s times, dialogue and dialectical reasoning have been regarded as crucial components of education. In the Middle Ages (Kretzmann et al. 1982), dialectics was together with grammar (linguistics) and rhetoric the ground of academic teaching, and argumentative tools (such as commonplaces, strategies and patterns of reasoning) were used as instruments for teaching and learning, and as part of the method for supporting philosophical theses. On the other hand, interest in argumentation studies is constantly increasing in the field of education. Argumentative, collaborative dialogues have been shown to be an extremely effective teaching strategy (Murray 1982; Glachan and Light 1982; Alexopoulou and Driver 1996; Baker 1999; Driver et al. 2000; Nussbaum 2008a, b) both for improving students’ critical skills (see Pera and Sahea 1991; Koballa 1992; Osborne 2010) and addressing background knowledge (Jucks and Paus 2013; Paus and Jucks 2012; Hewson 1992; Limón 2001), including misconceptions and the roots of deep disagreements. This principle is at the basis of the idea of science as an argument, i.e. the view that science education should address “the appropriation of scientific discourse”, constructing students’ knowledge through argumentative interactions (Kuhn 1991, 1993, 2010; Garcia-Mila et al. 2013). Through the use of arguments, students have been shown to acquire not only the knowledge of scientific concepts, but the grounds on which such concepts are based (McNeill and Krajcik 2008; Sandoval and Millwood 2005), revising their prior knowledge (Cross et al. 2008).

This paper is principally aimed at addressing the problem of providing a methodological tool for analyzing students’ arguments, and detecting what can count as a better or more sophisticated argumentative strategy (Erduran 2007). By developing a coding system that captures from both a dialectical and content point of view the nature of an argumentative move, it is possible then to design and propose possible empirical activities aimed at improving students’ argumentative skills. Our theoretical proposal is based on the combination of the two distinct traditions of argumentation studies and educational research. Insights provided by recent works in the field of argumentation have been used to design, code, and analyze an empirical study whose purpose is to assess qualitatively how students’ argumentative skills can be influenced through an expert–peer interaction in which the expert used specific strategies. The theoretical argumentative background provided the framework for selecting the argumentative strategies that directly affect the interlocutors’ replies.

In order to understand the argumentative effects of expert–peer interactions, we have designed two distinct coding systems, one aimed at capturing the dialectical structure of the argumentative dialogue, and the other focused on the content of each dialogical move. In the first case, we combined a functional coding scheme developed in the field of education (Kuhn and Udell 2003 and Nussbaum and Schraw 2007) with the dialectical principles pointed out in argumentation studies (Walton et al. 2008; Walton and Krabbe 1995; Krabbe 2003). The content level of the moves was coded based on the types of argument described in the argumentation scheme approach (Walton, Reed and Macagno 2008), which can be considered as a further development of Toulmin’s argument pattern. The use of this twofold coding framework is aimed at (1) examining qualitatively at which level a specific dialectical behavior adopted by the expert can lead the student to modifying his behavior, and (2) describing how such more sophisticated argumentative behavior can unveil deeper misunderstandings or disagreements. This structural and content analysis reveals how students can be led, by means of specific types of attacks, to engaging in dialogues of a deeper level, also called meta-dialogues, aimed at bringing to light the background beliefs underlying a specific position or claim (Leitão 2000; Nussbaum 2008b).

1 Dialogues, Argumentative Behavior, and Background Knowledge

Argumentation is a dialogical activity grounded on two crucial principles, i.e. common beliefs and defeasibility. Arguments are based on premises that are not necessary, but only commonly shared, such as values, presumptive causal relations, or commonly accepted definitions. Given the endoxical nature of their premises, arguments are inherently defeasible, i.e. subject to default should one of the premises be challenged or refuted (Nussbaum and Edwards 2011: 444). The essential relationship between background beliefs and arguments, and the defeasible nature of argumentative reasoning make argumentation an activity of fundamental importance for unveiling and addressing background knowledge and misconceptions.

The crucial role of argumentation in teaching has been underscored in many studies (see Pera and Sahea 1991; Osborne et al. 2004; Koballa 1992; Sandoval and Millwood 2005). Teaching can be considered as an argumentative dialogue as its basic purpose is to modify and develop the students’ “private understanding,” showing the limits thereof and building on it in order to account for new phenomena (Simons et al. 2001: 7; Chi and Roscoe 2002). One of the crucial instruments used for achieving such a goal is the use of arguments. Students’ arguments can be encouraged in order to bring to light the background beliefs on which they are based. Such arguments and the underlying background knowledge can become the issue of further argumentative moves, directed to providing reasons showing their incompleteness and supporting the scientific viewpoint. For this reason, argumentation tools are fundamental for both enhancing the students’ critical thinking skills, and improving classroom interactions, so that students’ prior beliefs can be better made explicit, addressed, and developed (Osborne et al. 2004: 995; Carey 2000: 13–14). The advances in argumentation theory can provide useful resources for improving argumentation in classroom interactions, by taking into account both the argumentative dialogical models (van Eemeren and Grootendorst 1984, 1992, 2004; Walton 1984; Walton and Krabbe 1995) and the argument structure (Walton et al. 2008).

These two dimensions of argumentative dialogue, i.e. the dialogue and the argument structure, can reveal distinct and interdependent aspects of argumentative behavior. On the one hand, the analysis of the different dialogical moves, i.e. speech acts aimed at a specific dialogical purpose (such as countering the interlocutor’s argument or advancing a viewpoint) can show the macro-structure of the speakers’ dialectical attitude. This analysis can mirror what type of argumentative discussion the interlocutors are engaged in, whether it is more focused on the opposition of incompatible positions (consisting in more frequent use of direct attacks), the search for an alternative and potentially acceptable viewpoint (corresponding to a more frequent use of arguments supporting an alternative conclusion), or on the investigation of the deep reasons supporting a certain standpoint (i.e. a more frequent use of undercutters). On the other hand, the argument structure can bring to light the distinct perspectives under which the issue under discussion and the interlocutors’ viewpoints (Walton 2006) are taken into account for supporting a conclusion. The same issue (such as military intervention) can be regarded under distinct complementary respects for the purpose of bearing out a certain position. For example, the same state of affairs can be taken into consideration from the point of view of its positive or negative consequences, or the values and rules that it involves, or the reasons that brought it about, or the way in which it can be classified and judged.

The two distinct types of analysis, focused on the two components of the argumentative behavior, i.e. the dialectical moves and the argument types, can shed light on the goal (and quality) of the students’ dialogues. A difference of opinion can be resolved by reaching an agreement on a viewpoint, which can correspond to the one defended by one of the parties or a new position encompassing the two opposing views. In both cases, at least one party needs to modify his attitude towards his viewpoint. In the first case, “one party has to be convinced by the argumentation of the other party of the admissibility of that party’s standpoint” (stronger condition) or the other party has to retract “his standpoint because he realizes that his argument cannot stand up to the criticism” (weaker condition) (Van Eemeren and Grootendorst 2004: 58). This latter option opens up the possibility of accepting the interlocutor’s or an alternative point of view. Dialogues, in order to be considered as closer to an ideal model of dialogical rationality (see for instance Van Eemeren and Grootendorst 1984, 1992, 2004), need to address the origin (or cause) of the disagreement by investigating the values and the classifications underlying the opposing positions. In this fashion it is possible to avoid pseudo-agreement or pseudo-disagreements (Naess 1966: 92–93) and direct the discussion to the “facts” (i.e. the subject matter that is controversial) and not merely to the words used (Aristotle, Topics 108a18–108a36).

The distinction between dialogues focused on the standpoints and the ones aimed at unveiling the deeper cause of disagreements is mirrored by the theoretical difference between the dialogical and meta-dialogical levels of dialogue (Krabbe 2003, 2007). The first level consists in moves directed at attacking the interlocutor’s viewpoint by opposing a contrary or an alternative viewpoint, which in turn needs to be defended when attacked. The second level can be described as a dialogue on the conditions of a dialogue, or a discussion on the grounds of a difference of opinions. The speaker challenges the possibility of a dialogue move, which in case of arguments amounts to the relationship between the premises and the conclusion. In other words, he challenges the fact that a premise can bear out the wanted conclusion because based on starting points that are unaccepted by or unacceptable for the hearer. For this reason, meta-dialogues usually bring to light a deeper dimension of a disagreement (Turner 2006; Fogelin 2005; Finocchiaro 2012: 40), consisting in conflicts of classifications, definitions, and values. Even though there is no correlation between a meta-dialogue and a specific argument type (dialogue and meta-dialogues can consist of different kinds of argument), the use of certain types of argument can reveal a specific way to address deep disagreements. For instance, a higher use of reasoning from classification shows the need of analyzing in depth a concept or the interpretation of a state of affairs (Chi 2005). Conceptual representations of words that represent central aspects of a learning topic can be highly different between learners (Paus and Jucks 2011; Jucks and Paus 2012). Resolving such differences during discussion results in deeper elaboration of the underlying conceptualization and is shown to have a pivotal influence on conceptual learning (Paus and Jucks 2012; Paus et al. 2012; Jucks and Paus 2013).

2 Analyzing and Developing Argumentative Behavior

The evaluation of argumentative behavior based on the dialectical level of argument questioning and attack constitutes the ground for advancing a method for improving learners’ argumentation quality by designing students’ discussions, which include the use of deeper dialectical moves. This proposal is grounded on two distinct approaches to dialogue and argumentation, i.e. empirical educational research and theoretical argumentative models.

In education, models have been developed to improve students’ reasoning and argumentative competence by teaching argumentative skills (Zohar and Nemet 2002) and guiding and modeling discussions among peers and between experts and students. Peer–peer discussions are directed in particular to improve collaborative reasoning (Clark et al. 2003; Andriessen 2006; Kuhn et al. 1997). Cognitive conflicts to be solved by discussants stimulate students’ argumentation, consisting not only in the production of arguments, but also in the evaluation, questioning, and countering thereof. In particular, in Anderson et al. (2001) and Kim et al. (2007) it was shown how students in collaborative reasoning discussions tend to copy one another’s argument strategies. In this sense, the empirical education research showed how discussants tend to learn by discussing with a partner using particular argumentative behavior. Expert–peer interactions have been modeled in order to improve students’ construction of arguments (Erduran 2007; Osborne et al. 2004) and encourage students’ reflection about their positions (Jiménez-Aleixandre 2007) and reasons (Zohar and Ben David 2008), developing meta-cognitive skills through encouraging them to reflect on their own reasoning and arguments (Zohar 2012).

The teacher’s guidance used to improve students’ argumentative abilities can be analyzed and developed relying on the dialectical and linguistic studies. From a theoretical perspective, an argumentative move leads the interlocutor to adapting contingently to the speaker’s argumentative behavior. The performance of a dialectical move, such as questioning a viewpoint or attacking a premise, opens specific possibilities to the interlocutor while blocking others. He is led to carrying out specific dialectical moves to reasonably contribute to the dialogue game (Ducrot 1972; Macagno and Walton 2014; van Laar 2014). In this sense, a student–expert dialogue can affect the student’s dialectical behavior because the student’s possible moves and countermoves are limited by the expert’s dialectical choices. In this fashion, the expert can guide the student to perform more “critical” dialogical moves and enter deeper (in the sense of aimed at uncovering and solving the deeper cause of disagreements) types of dialogues. From an empirical point of view, previous studies showed that teacher-guided discussions were more efficient in “attaining higher levels of reasoning and higher quality explanations” (Hogan et al. 1999: 379; Kuhn and Crowell 2011).

In particular, we observed the quality of peers’ argumentation with former interaction with an expert,Footnote 1 i.e. an arguer that uses more sophisticated (or rather deeper) argumentative (in this case dialectical) strategies. The purpose was to illustrate how discussions with experts can improve the interlocutors’ argumentative behavior by enhancing a deeper understanding of the reasons underlying a position, and affecting the dialectical analysis of an issue. Our goal is to show qualitatively how and whether an expert–peer dialogue, by guiding the student to focus on the deep causes of the disagreement, can influence his dialogical behavior in further dialogues with a peer. This empirical study suggests to two distinct and interrelated questions, one empirical and one theoretical. From an empirical point of view, addressed quantitatively in Paus et al. (submitted), this study opens up the possibility of advancing a method for improving—at least in the short term—the students’ argumentative behavior. From a theoretical point of view, tackled in this paper, the problem is to establish whether the expert, by adopting a more critical argumentative attitude (attacking the premises and revealing the grounds of the student’s position), can induce the student to imitate the same dialectical strategies and focus more on the deeper issues presupposed by the conflict of opinions.

3 A Coding System Detecting Argumentative Skill Development

The idea of “argumentative strategy,” developed in Anderson et al. (2001) and Kim et al. (2007) for analyzing the snowball effect, tended to merge two levels that are theoretically distinct, i.e. the structure and the content of an argumentative move. Under the label of “strategy” fall both types of argument, defined based on their content (such as from consequences), and kinds of dialectical moves, such as questioning or rebutting. From a theoretical perspective, these two levels need to be distinguished, as they affect differently the interlocutor’s behavior. As mentioned above, dialectical moves, intended as speech acts aimed at achieving a specific reply for the purpose of supporting or weakening a point of view, influence directly the interlocutor’s behavior by limiting the paradigm of his possible replies. On the contrary, from a theoretical point of view this effect is not brought about on the content level. The refutation of a move or the questioning thereof can be performed by resorting to various kinds of arguments or types of reasoning, which are not determined by the speaker’s argumentative choices. For this purpose, we used two distinct types of coding schemes. The first scheme is focused on the types of dialectical moves (functional level), while the second one takes into account the types of argument used (content level), in order to bring to light the nature of the positions and background knowledge that is debated.

3.1 Developing a Coding Scheme—Empirical Study

The development of the two distinct schemes was based on a sample of 16, 11 to 13-year-old, 6th to 8th-grade students (approximately half female) attending a public middle school in New York City. They were all participants in an ongoing twice-weekly curriculum in argumentation that the school had adopted as part of its regular curriculum. It focused around electronically conducted peer dialogs between a pair of students who held one position on a social issue and a succession of pairs of their peers who held the opposite position. Participants came from the same population as that described in earlier publications regarding this curriculum (Kuhn and Crowell 2011; Crowell and Kuhn 2014; Kuhn et al. 2013). The school is an academically challenging urban public middle school in an ethnically diverse low-to-middle-income neighborhood. Eighty percent were Hispanic or African-American, and 60 % qualified for free or reduced-price lunch.

For the purpose of this study, three out of six students’ electronic dialogs on a single topic were chosen for investigation. In the second of these three dialogs, an adult expert (i.e. an arguer instructed to use specific argumentative moves) was substituted for an opposing peer pair, unbeknownst to the participants. Based on these dialogs we will now illustrate how the interaction with an expert can influence students’ argumentative behavior on a functional and a content level.

3.2 Coding Scheme—Functional Level

On the level of the structure of the dialectical moves, our analysis was focused on student pairs’ use of strategies to attack the interlocutor. Strategies of attack are based on three distinct ways of refuting a viewpoint: (1) supporting an alternative and incompatible viewpoint (Counter A); (2) rejecting directly the interlocutor’s viewpoint, showing its falsity or unacceptability (Counter C) (Felton and Kuhn 2001); (3) attacking by questioning or rejecting one of the premises on which the interlocutor’s argument is based (Underminer). On this perspective, Counters A and Counters C represent a counterargument and a defeater (Pollock 1974, 1987), while underminers include the so-called undercutters, i.e. the attacks against the inferential link between premises and conclusion, and the questioning or rejecting of the premises that support the conclusion by supporting their falsity or asking critical questions (Walton 2006; Walton et al. 2008; Nussbaum and Edwards 2011). This distinction allows one to identify the dialogical moves that uncover the interlocutors’ background beliefs. Underminers, by focusing on the reasons underlying a conclusion, can reveal deeper assumptions that are the basis of an argument, such as values, commitments, or criteria of classification. To code the dialogs, only content-related contributions serving to discuss the topic (such as “if they continue to misbehave, they should be expelled”, “they could be homeschooled by a tutor”) were taken into account. The segmentation was based on so-called dialog units. Note, that a dialog unit was not necessarily similar to a complete message, because in online dialogs a message is likely to include contributions of various length, structure and content. Consequently, a dialog unit can be a complete message, but can also be only one subpart of it (Jucks and Paus 2013; Asterhan and Schwarz 2009). Interrater agreement for roughly 50 % of the dialogs was very good with Cohen’s Kappa κ = .91 (Krippendorf 2004). See Table 1 for the functional coding scheme.

Table 1 Categories of the functional coding scheme
Full size table

3.3 Coding Scheme—Content Level

From the point of view of the types of argument used, the analysis has been based on the theory of argumentation schemes developed in Walton, Reed and Macagno (2008), in which the stereotypical patterns of argument are formalized. Argumentation schemes represent the combination of the material (semantic) link between premises and conclusion (such as cause-effect, authority, or consequences) and the type of reasoning used (such as modus ponens, abduction, or induction). In particular, six types of argumentation schemes have been chosen as the most adequate for the kind of task (supporting a decision) and topic (an ethical issue, expelling disturbing students from school) (Macagno and Walton forthcoming). The schemes were the argument from consequences and the argument from practical reasoning (schemes used for supporting a decision), the argument from values and the argument from rules (justifying an ethical or normative viewpoint), the argument from best explanation (tracing back a behavior or event to the most reasonable cause), and the argument from classification (supporting the categorization of a behavior, event, or state of affairs in a certain fashion). Other argumentation schemes that have been used, in particular analogy and example, have been placed under the category of classification, since they have been used only for classifying a specific behavior. Here, again the data coding was solely based on the content-related dialog units. As for the functional scheme, agreement between raters for roughly 50 % of the dialogs was good with Cohen’s Kappa κ = .87 (Krippendorf 2004). See Table 2 for the content coding scheme.

Table 2 Categories of the Content Coding Scheme
Full size table

Critical questions have been codified according to the schemes that they are undermining, as they belong to the same argument structure. For instance we consider the following argument attacked by using critical questions (Table 3).

Table 3 Critical questions to attack arguments
Full size table

The theoretical distinction between more and less sophisticated argumentative strategies and the design of the two interrelated coding schemes can be used to develop specific interventions aimed at improving students’ argumentative behavior. In the section below, the qualitative results of the empirical study mentioned above will be presented, in order to provide the first hints for possible classroom activities.

4 First Insights into Students’ Argumentative Behavior

The two coding schemes are aimed at detecting and analyzing the function of the dialogue units used to attack the interlocutor’s move, and the argumentative content of the students’ move. The purpose was to provide a method for detecting more sophisticated dialectical behavior, and describing how the dialectical moves are related with arguments. In this section we will illustrate how the coding schemes introduced might help us to better understand students’ improvement in argumentative behavior after being exposed to an interaction with an expert. The goal is to provide a qualitative analysis of the prototypical dialogues before, during, and after an expert interaction, to examine the changes in dialectical behavior from a functional and content perspective.

From a quantitative point of view (discussed in detail in Paus et al. submitted), a noticeable change can be noted in the types of attack to the interlocutor’s point of view before and after the expert interaction. The students tended to take into better account the interlocutor’s position by attacking it, either by countering it directly (from 37 % to 47 %) or undermining it (from 16 to 22 %). On the contrary, they opted much less for the mere opposition of an alternative point of view (counter-alternative) (from 17 to 5 %). This type of behavior was mirrored from a content perspective by the types of argument used by the students. The practical reasoning, an argument preferentially used to provide an alternative solution to a problem, decreased (from 23 to 19 %), while the use of values (which are mostly used in solving deep disagreements, see Macagno and Walton 2014) increased (from 17 to 25 %). For the other categories of argument the difference was <3 %. The changes in behavior that these figures hint at will be illustrated qualitatively by analyzing the dialogues that are most representative for the type of interaction.

4.1 Prototypical Peer–Peer Pair Dialogue (PP1)

The following peer–peer pair dialogue shows both the structure of the moves and the types of arguments that have been noticed to be most frequently used between students (Table 4). The students choose to attack the conclusion of the interlocutor directly, either by drawing the negative and potentially unacceptable consequences (moves no. 4, 6, 8, 12) or proposing an incompatible alternative (moves no. 2, 3, 5, 7, 11). Only in one case (move no. 13) the first student (ex2) undermines the interlocutor’s argument by rejecting a premise upon which the whole argument rested.

Table 4 Example of an expert–peer dialog
Full size table

4.2 Prototypical Peer–Expert Pair Dialogue (PEX)

The type of dialogical behavior is different in peer–expert interactions (Table 5). The expert questions the principles (the values) at the basis of the student’s viewpoint (based on rules), pointing out that there are other possible causes of Doris’ and Roger’s misbehavior, which can result in other more reasonable and useful ways to solve their problems (moves 4, 6). The student is brought to reflect upon the interpretation of the kids’ disruptive behavior, and on the responsibility of schools in addressing their problems (moves 5, 7). The expert, in this fashion, shifts the dialogue to an analysis of the way that the school has handled the kids’ behavior and the values (the social values in this case) at the basis of the choice of expelling Doris and Roger (moves 8, 10, 12).

Table 5 Example of a peer–expert dialog
Full size table

This interaction shows how the argumentative strategy of the expert is focused on unveiling the reasons at the basis of the student’s position, i.e. the interpretation of the state of affairs under discussion (the reasons of the kids’ behavior) and the values underlying the expulsion. By undermining the interlocutor’s arguments, the expert addresses the roots of the disagreement.

4.3 Prototypical Second Peer–Peer Pair Dialogue (PP2)

The peer–peer interaction that followed the dialogue with the expert shows a strategy that is different from the one pursued in the first peer–peer interaction (Table 6). The students do not attack the conclusions of the interlocutors’ reasoning directly, drawing consequences or proposing alternatives. Instead, they challenge the principles underlying the viewpoints, pointing out internal contradictions. This dialogue is based on the values underlying the positions (expelling the students or punishing them in other ways) (moves 2, 7, 11, 12, 13) and the interlocutors’ deep commitments about them. This latter issue in particular is discussed in a dialogue of a different level, focused on the grounds of a position (14–22).

Table 6 Example of a peer–peer dialog after expert interaction
Full size table

This latter dialogue shows the dialogical difference made by the use of underminers. Underminers can shift the dialogue to a deeper level, leading the interlocutor to addressing the reasons of a specific viewpoint. Such dialogues could be crucial for solving deep disagreements or misunderstandings.

5 Underminers, Meta-dialogues, and Background Knowledge

As noticed in the dialogues above, underminers can be used to open a dialogue of a different level, called “meta-dialogue” (Krabbe 2003; Walton 2007). Meta-dialogues are aimed at discussing not the acceptability of a viewpoint, but rather the conditions of a dialogue or, in the specific cases analyzed, the conditions of an argument. Meta-dialogues can be focused on how a premise can support a conclusion, i.e. the material link and the respect under which a premise is relevant for the conclusion (Govier 1985; Macagno 2008; Macagno and Zavatta 2014). The speaker for instance can attack the relationship between a characteristic of the state of affairs advanced as a reason and the material relation linking it to the conclusion. This attack can open a dialogue on the acceptability of a specific dimension the argument itself. In this sense, meta-dialogues and in many cases underminers can play a crucial role in the development of metastrategic knowledge, i.e. the awareness of the thinking strategies that are used (Zohar and Ben David 2008: 60; Zohar 2012: 199). The students, by assessing, attacking, and discussing the acceptability of an argument or the premises thereof, can acquire awareness of the thinking mechanisms and argumentative reasonableness. For instance, we consider the following excerpt (Table 7), in which the meta-dialogue becomes a dialectical instrument for assessing the reasonableness of the classification of a state of affairs and the application of a rule:

Table 7 Meta-dialogue 1
Full size table

In this dialogue, the first student (ex3) supports his conclusion based on the principle of inference drawn from the rules of the school (after a certain number of warnings, a student shall be expelled). The second student (ST1) attacks the relationship between the premise and the principle of inference, pointing out that the conclusion can be drawn only depending on the type of warnings. This attack (questioning) opens a meta-dialogue on the relationship between the premise and the conclusion of the argument. The first student defends the principle of inference on which he grounded his argument (the quantity of warnings results in expulsion), which is then rejected by a different rule (the quality of the warnings determines the punishment), which ultimately leads to the undermining of the whole argument (if the type of warnings has not been specified, you cannot punish the students in the most severe way).

In the following case, the meta-dialogue is focused on an argument from values (Table 8). The student (ex7) claims that the two students Doris and Roger should be expelled based on the rule that after a certain number of warnings they should get a serious punishment. The expert attacks the argument and proposes alternative ways to solve the problem (meeting with the parents to understand and fix the root of the students’ problem). At move no. 9, the student insists in focusing on the bad behavior (they are not good persons) and the expert starts a meta-dialogue on the ethical principles at the basis of the rules of the school (move 10).

Table 8 Meta-dialogue 2
Full size table

The meta-dialogue directs the student to take into account the ethical principles behind expulsion and punishment. As a consequence, he accepts the value advanced by the expert but attacks the applicability thereof to the specific case, as the students do not want to be helped (move 11 and 13).

Meta-dialogues can be focused on different types of argument, such as classifications (meta-dialogue 1) or values (meta-dialogue 2). In (Table 8), at move 7, ex7 opens up a meta-dialogue on the reasonableness of the interlocutor’s practical reasoning, which leads to a dialogue on the values underlying a decision (moves 10–11) and the interpretation of the state of affairs on which they are discussing about (moves 12–13).

6 Discussion

6.1 Conclusion

This paper addresses from a qualitative point of view two interrelated issues: how to detect, describe, and assess the argumentative behavior of students in classroom discussions, and how to intervene in order to improve their argumentative abilities. In this sense, we tried to provide a method for evaluating dialogical moves according to argumentative criteria, in order to distinguish the more sophisticated and effective strategies from the others. This theoretical proposal has been used in a practical fashion, to propose a possible classroom intervention.

The first question was tackled by introducing a twofold coding scheme aimed at capturing the argumentative behavior of students’ interactions from both a functional and content perspective. The interrelated use of the two coding schemes shows how it is possible to merge the more functional approaches (developed mostly by Kuhn and her collaborators) with the more structural and argument-oriented perspectives based on the Toulmin Argument Model. How shown by the qualitative analyses of students’ dialogues, the functional and the content schemes can code and bring to light different types of argumentative moves and describe them according to their role in addressing the other’s perspective and influence the process of belief change. In particular, this paper underscored how specific dialectical moves can be used to address and discuss background knowledge, i.e. the deeper grounds of a specific viewpoint. In this fashion, argumentative behavior can be examined and assessed according to its dialogical function, i.e. to address the other’s point of view and guide the interlocutor to a change of perspective. The deeper and more sophisticated argumentative moves can be detected through the functional scheme, while the content scheme discloses the dimension of the viewpoint that is discussed or attacked in the dialogue. From a qualitative perspective, we have shown how the two levels of analysis can be used in combination to bring to light the nature of the dialogues on deep disagreements, which we called meta-dialogues. This account shows how the same types of argument can be used at different levels with distinct effects. For instance, arguments from values can be used to attack a position by opposing to it a different value; however, only when an argument is undermined and the hierarchies of values are discussed it is possible to address the root of a disagreement.

These theoretical insights shed light also on the second issue, concerning the possible interventions aimed at improving students’ argumentative skills. This study provides first hints at possible activities that can be used to influence students’ dialogical behavior. We have shown from a qualitative point of view how interactions with an expert—who uses specific and more sophisticated types of attack—can affect the students’ argumentative behavior also in later dialogues. In this sense, this study points out the possibility of transferring the skills of argumentation, and in particular the strategies focused on addressing deeper misunderstandings or disagreements, through practice, and not simply teaching (see Zohar and Nemet 2002).

6.2 Limitations and Implications for Further Research

The results of this study provide a promising approach to foster argumentative skill development in the classroom. Curricula (see Kuhn et al. 2014) based on the approach of argumentation as a dialogic process (as distinct from argument as a product, see Kuhn and Udell 2007) suggested integrating short interactions with a more sophisticated interlocutor. Experiencing a real (skilled) interlocutor—instead of being taught argumentative strategies by instruction—enables them to internalize the dialogic frame and helps them understand the purpose of argumentation and argumentative writing. In this vein, such “natural” learning approaches could be of high interest for developing methods for tutoring or designing remedial lessons regarding motivational aspects with students becoming an active part of the teaching process (Ames 1992).

The practical outcomes of this study are grounded on specific concepts that can be generalized. In particular, this work has taken into account a particular type of expert (or rather an interlocutor using more sophisticated argumentative strategies), a particular type of population, and particular types of topics that have been treated in argumentative dialogues. These three defining characteristics of the study, however, can be encompassed by more general considerations that can apply to a much broader idea of expert, to a much wider population, and to different types of subject matters.

Concerning the first limitation, we noticed in the paper that the students’ argumentative behavior was influenced by the interlocutor’s moves. In this specific case, the experts were Ph.D. students, i.e. adults with a specific training in education. However, these results suggest that it is possible to generalize the idea of improving students’ argumentative skills through interactions with a generic expert. Teachers or tutors can focus their argumentative dialogues on questioning the premises of a position, i.e. asking the critical questions that could undermine an argument (Walton et al. 2008), until attacking the classifications and the definitions at the basis thereof. Our argument structure analysis provides a guideline for addressing specific types of attacks (the critical questions) related to the kind of arguments used by the interlocutor. By identifying the type of argument, arguers can retrieve the specific type of underminers that can be used to attack the interlocutor’s reasoning (Walton 2005).

The findings presented here are limited to the population of middle school students whose argumentative skills are rather weak and who might preferentially benefit from a discussion with a partner that uses deeper or more sophisticated argumentative strategies. However, the specific dialectical strategies analyzed in this paper can also be used to make arguers aware of the types of moves that they are performing and the effects that they can elicit. Future research should look at other populations (such as university students) to find out whether and how such an intervention also works for more skilled arguers.

Concerning the last limitation, this study was conducted by taking into account ethical issues (punishments in schools, war intervention) in order to point out the dialectical dimension of the discussions. However, the outcomes of this study provide insights that can be applied to other educational contexts. Distinct strategies of refutations, and in particular underminers (Damer 2001), can be used to induce the students to evaluate the previous scientific conceptions on which their arguments were based. For instance, undercutting arguments by attacking the definitions of crucial concepts (such as the idea of “temperature” in discussions about thermodynamics) can be an effective instrument for guiding students to restructure their view on and their understanding of a specific issue (Konstantinidou and Macagno 2013; Nussbaum and Sinatra 2003; Cross et al. 2008).

The last global consideration concerns the duration of the effects of the argumentative strategies analyzed in this study. This work is based on a short intervention, which, however, was shown to already elicit changes in the argumentative behavior of middle school students. This effect raises the question of whether such effects would remain over a longer time period. Will students still take advantage of such an intervention after a couple of weeks or even a year? Would their argumentative behavior improve even more, if they participated in expert-dialogs on a regular basis? A possible direction of future works could be the investigation of the possible effects of this kind of argumentative interventions, i.e. the long-term effects on argumentative skills (increased awareness of argumentative strategies), and the short-term learning effects (conceptual change) in different educational domains.