Arguing in Direct Democracy: An Argument Scheme for Proposing Reasons in Debates Surrounding Public Votes

We develop a novel argument scheme tailored to debates surrounding public votes on a state action. It can be used to propose reasons for voting “yes” or “no” and allows for natural reconstructions of such debates. These reconstructions are of particular use to voters trying to weigh the pros and cons of the proposed state action. The scheme for proposing reasons helps answering two questions: What changes will the proposed state action bring with it? And are these changes good or not? Using the scheme, we derive a three-layered structure of public debates surrounding votes that aides in constructing systematic and comprehensive overviews. The scheme for proposing reasons also lends itself to concise summaries for voter information and can be incorporated into more general deliberation dialogues.


Introduction
Every now and again, the voting public in a democracy can decide directly on some state action. Typically, this choice takes the form of a referendum or a popular initiative. There, some possible state action is proposed to the public and carried out if it receives enough votes (most often a simple majority) in its favour.
In case of a controversial state action, its proponents engage in a public debate with the opponents in order to convince the voting public of their respective view. The final decision on the action is then taken by the accumulation of the individual votes cast by the members of the public.
The public debate plays an important role in the individual decision made by the voters. As Kriesi notes in a comprehensive study of Swiss votes, particularly the arguments exchanged by the proponents and opponents are crucial to the way the public casts its votes (Kriesi 2005). This function of public debates is mirrored in Kock's more normative approach where he states that "political debate should function as input for citizens' deliberation" (Kock 2018, p. 493).
In this paper we develop an argument scheme that aides in systematic reconstructions of public debates surrounding votes. Using this scheme, one can provide a clear overview of these debates that facilitates critical engagement with their arguments. This way, we aim to enhance the utility the voting public can get out of the often complicated and convoluted debates that surround public votes.
We will begin with some preliminary comments in Sect. 2 in order to briefly introduce the Swiss political context, which will feature as the basis for our reconstructions, and say a few words about how we approach things. Section 3 will then examine other approaches and show how ours differs from them. As the concept of reasons for voting "yes" or "no" will be crucial for our scheme, Sect. 4 is dedicated to examining those, and Sect. 5 will then present the centrepiece of this paper: the scheme for proposing reasons itself. Through this scheme we can then induce a three-layered structure, presented in Sect. 6, that allows us to capture large parts of public debates surrounding votes. The use of our reconstructions is illustrated in Sect. 7. We end in Sect. 8 with discussing the application of our scheme in deliberation dialogues.

3 2 Preliminaries
Before we can move on to discussing our and others' approaches, we give some preliminary comments on Swiss politics and the general aims of our reconstructions.

Background on Swiss Politics
In developing the scheme for proposing reasons, we focus on procedures and argumentation found in Switzerland. Switzerland has a rich history of direct democratic elements [see e.g. Vatter (2020)] and holds regular votes (usually four times a year) where the public can decide on changes of the laws and of the constitution.
Thus, in the context of these votes, the proposed state action is to change a law or part of the constitution. The voters are then, in most cases, presented with a binary choice of whether to accept or reject the proposed change, that is, to vote "yes" or "no". 1 In order to assist the voters in deciding how to vote, the federal council distributes a little red booklet for each vote called the Explanations of the Federal Council [henceforth: Explanations, Rigotti and Greco (2019)].
Apart from some general information about the proposed change in law or of the constitution and the specific wording of the change, the Explanations also contain a short summary of the public debate. Each side of the debate (one of them being the Federal Council and the Parliament) receives approximately 1.5 pages to present their arguments in their own words.
Throughout this paper, we will use these argumentative texts in the Explanations as our basis for reconstruction. This allows us to reduce the vast public debates spanning numerous forms of media, from television formats over newspaper articles to media conferences, to a manageable size without distorting it. As both sides of the debate have equal space and are equally free to make their case, we take the account provided by the Explanations to be adequate. And as this booklet is one of the most important sources of information for the voters (Golder et al. 2021), it has proved useful for various studies over the years [see Schröter (2022a) for an overview].

Aims of the Reconstruction
We distinguish roughly two aspects of public debates surrounding votes. On the one hand, it needs to be determined what reasons there are for voting in favour or against a given proposal. On the other hand, these reasons need to be weighed against each other. Where, in a first stage, reasons such as "voting yes will promote Swiss values" might be proposed, it needs to be discussed, in a second stage, how important it actually is to promote Swiss values.
In this paper we focus on the first aspect of debates surrounding public votes. Our general aim is to propose a method for providing overviews of the reasons and their justifications proposed in such debates. This gives us a first goal: To give an account that is true to the debate. The resulting account should not only be theoretically interesting, but also facilitate fruitful critical engagement with the arguments. Hence, there is also a second goal: To give an account of the debate that presents it in a way that is useful to the voters.
In accordance with Kriesi's and Kock's claims, we understand the purpose of public debates as to put forth and critically discuss various aspects of the proposed state action in order to assess its virtues and shortcomings. In a sense, then, these debates propose reasons for voting "yes" or "no". The individual voters can use such debates to consider the proposed reasons and check whether they agree with them or not. Thus, in providing a useful reconstruction of public debates surrounding votes, we will focus on the proposed reasons and their justifications.
Roughly, we claim that a reason for voting "yes" or "no" results from a change in the world brought about by the proposed state action. If the change is deemed positive by the voter, it will result in a reason for "yes". A negative change originating from the action will, accordingly, result in a reason for voting "no" (see Sect. 4).
We see two aspects for which a reconstruction focusing on such reasons can be useful to the voters: First, we think that these reasons provide the material that a yet undecided voter can weigh in order to reach a decision on how to vote. Our reconstruction delivers these reasons and their justifications. Second, our approach facilitates a critical examination of these reasons. Tying rhetorically laden buzzwords to concrete changes might, for example, turn them into something accessible to critical questions and reflection.
It is important to note that our two goals for reconstruction, being accurate and being useful, will influence the final product in different ways [see Mohammed (2016)]: Being accurate requires that one does not invent arguments or distorts the ones given to a point where they become different arguments. However, if the reconstruction should be useful, that is, highlight the proposed reasons for voting "yes" or "no", one might need to change the arguments to some extent. Our account has to work with both these goals and we hope that the reconstructions we propose strike an optimal balance.

Characteristics of Our Account
Using argument schemes to reconstruct public debates surrounding votes is not in itself a novel approach. Even the Explanations have already been analysed using argument schemes. However, the scheme for proposing reasons that we will develop exhibits three features that distinguish it from the schemes used in these analyses: (1) It makes the inferential structure of the arguments clear. Further, (2) instead of a call to action such as "Vote 'yes'!", it concludes with a reason for voting "yes" or "no". Finally, (3) our scheme focuses not on a goal or outcome, understood as some future circumstance, but on a change of circumstances from the present to the future.

Inferential Structure
A common approach stemming from the rhetorical tradition is to take the statements from the argumentative texts in the Explanations and assign them a fitting topos [see e.g. Reisigl and Wodak (2009)]. For instance, sentences such as "The moral obligation of developed countries to the developing countries is to create such an environment which [...] stimulates economic development of the less developed countries." can be assigned to the topos of moral duty (Reisigl and Wodak 2009, p. 111).
This can lead to interesting results, as Klein shows in his analysis of the Explanations (Klein 2018). He uses his own topical pattern for political argumentation (Klein 2000(Klein , 2019 and orders the statements in the texts accordingly. A statement like "The federal constitution guarantees freedom and basic rights" would, for example, be labelled as a topos of principles (Klein 2018, p. 100). 2 Analyses like these can be useful in bringing to the surface the flavours and types of statements put forth in a debate, especially so in cases of subtle hate speech. In that sense, it helps seeing what is really going on so that something can be done about it.
However, it is not clear how the statements listed according to their topoi come together to justify the conclusion of voting "yes" or "no". While the listed statements do indeed seem relevant for the conclusion, the inferential structure is not visible when the argument is presented as such.
In contrast, we aim to provide an argument scheme with a clear inferential structure. That is, it should be clear how the conclusion follows from the premises, so that it is clear to the voter what they are committed to if they accept the premises.

Reasons as Conclusions
According to our view, the most important task of the argumentative texts in the Explanations is to provide the voter with reasons for voting "yes" or "no". The final recommendation, or call to action, of these texts, "Based on all these reasons we recommend accepting/rejecting the proposal", indicates that the individual arguments are supposed to be taken together in order to justify a vote. That is, the individual arguments provide reasons which are then taken together to support a voting recommendation. Accordingly, our scheme concludes with a reason for voting "yes" or "no".
This contrasts with most argument schemes that have been, or could be, used to analyse the Explanations, as all of them conclude directly with a call to action. Prominent accounts are offered by Rigotti and Greco (2019) as well as Schröter (2020Schröter ( , 2022aSchröter ( , 2022b. Rigotti and Greco draw from their Argumentum Model of Topics, and find that an argument often used by the federal council corresponds to their locus from termination and setting up. Its characteristic premise (the maxim, in their terms) is: "If something is predominantly good, it should not be terminated (despite its minor defects)" (Rigotti and Greco 2019, p. 309). Such an argument then concludes with "The current regulation should not be terminated"-in other words a call to (in)action. 3 2 His analysis concerns a vote on banning the building of minarets in Switzerland and in full looks as follows: [ (Klein 2018, pp. 99-100), our translation]: Topos of data -Rising number and increasing political-legal demands of muslims in Switzerland.
-Democracy as a means to the end of achieving the political-societal power of islam and minarets as a weapon in the fight for it.
-The minaret is not mentioned in the Koran.
Topos of valuation -The minaret has no religious function.
-It is a symbol of the societal-political claim to power of Islam.
-It is the spearhead of the political islamification.
Topos of principles -The federal constitution guarantees freedom and basic rights.
-The sharia stands in inextricable contrast to that.

Finaltopos -No to the islamification of Switzerland.
Conclusion -Yes to the prohibition of minarets.
It is worth noting that all these statements have been published in unattenuated form as part of the official federal voter information (Schweizerische Bundeskanzlei 2009). 3 A full example looks as follows (Rigotti and Greco 2019, p. 310

):
Endoxon: That contracting parties and social partners can shape tailor-made solutions means, in general, that regulation on holidays has proven effective despite possible minor defects.
Datum: With the current regulation on holidays, contracting parties and social partners can shape tailor-made solutions.
First conclusion: The current regulation on holidays has proven effective despite possible minor defects.
Maxim: If something has proven effective, it should not be terminated despite possible minor defects.

Final conclusion:
The current regulation on holidays should not be terminated.

3
Similarly, Schröter finds that more than 90% of all arguments in the Explanations that speak directly in favour of voting "yes" or "no" fit the argument scheme for pragmatic arguments [the scheme stems from Perelman and Olbrechts-Tyteca (1969)] (Schröter 2022a, p. 61). There, some effect of an action is framed as positive or negative, from which a call to action for voting "yes" or "no" is inferred. 4 Apart from the schemes that have already been applied to the Explanations, the scheme for practical reasoning suggests itself to be used for our purposes. After all, it is supposed to capture reasoning about what to do [see Audi (2006)], just as the voter is reasoning about how to vote.
While there are various versions of this scheme [see Fairclough and Fairclough (2012), Lewiński (2018), Macagno and Walton (2018)], they all share the call to action in the conclusion [see Lewiński (2021) for a discussion]. 5 We suggest locating the conclusions of the arguments in the Explanations at the level of reasons instead of the call to action based on several grounds: Apart from the function of the already mentioned final recommendations in the texts, we should also consider that the individual arguments are not put forth in a vacuum, but rather as part of a whole debate. Thus, a single argument can hardly be sufficient for establishing a call to action on its own, as such calls depend on a balance of reasons. While reasons do occur in the arguments, reconstructing them as Rigotti and Greco as well as Schröter suggest gives off a false impression. Namely, such a reconstruction suggests that accepting the premises commits one to accepting the call to action, that is, to voting "yes" or "no". This is far too strong.
What a call to action as the conclusion does, in essence, is to mark the reasons in the premises explicitly as reasons in favour of or against the proposed state action. However, it can happen that what for someone is a reason in favour of a proposed state action, is a reason against it for someone else. Say, some proposal benefits the rich. This benefit can be a reason in favour of the proposal for a rich person (presumably concerned with getting richer), while a poorer person, concerned with the rich not getting richer, will take the same reason as counting against the proposal.
This means that if we consider what is important for the voters, rather than the connection of a reason to a direction (i.e. in favour of or against) as depicted by the typical reconstruction, the reasons expressed in the premises and their justification takes centre stage. Thus we need a change in perspective and shift our attention further up the argumentation and concentrate on reasons for voting "yes" or "no".

Focus on Change
The scheme we propose is not the only one put forth in the literature that has a clear inferential structure and focuses on reasons. Macagno and Walton have constructed an argument scheme from positive/negative value that also combines these two characteristics (though not in the context of votes) (Macagno and Walton 2014, p. 52;Walton et al. 2008, p. 321): (1) The state of affairs x is positive/negative as judged by agent A according to value V (value judgement).
(2) The fact that x is positive/negative affects the interpretation and therefore the evaluation of goal G of agent A (if x is good, it supports commitment to goal G/if x is bad, it goes against commitment to goal G).
(3) The evaluation of x according to value V is a reason for retaining/retracting commitment to goal G.
Roughly, we can understand a goal as an aspect of a state of affairs that is judged to be desirable based on some value. A state of affairs is here understood as a comprehensive description of what is the case at a specific time (Bench-Capon and Atkinson 2009, p. 49;Atkinson et al. 2005, p. 155). This scheme then expresses how such a judgement comes about. First, one evaluates a state of affairs as positive or negative (1). This is then connected to a specific goal (2) and taken to provide a reason for retaining or retracting it (3).
Much of the heavy lifting is done by premise (1), and much depends on how exactly one evaluates a state of affairs. Take someone who is concerned with achieving a just society and consider their stance towards the introduction of paternity leave. It is arguable that its introduction, while making our society a bit more just, still leaves it unjust as a whole (one only has to consider that much of our wealth is built on the cheap labour of third world countries). Thus, the state of affairs with paternity leave would be evaluated as unjust and thus as negative. But then, according to premise (2), 6 they would have a reason to retract commitment to introducing paternity leave, even though it would make our society a bit more just. 4 For example (Schröter 2020, p. 138): (1) If the effect/consequence is to be refused, then the cause is to be refused.
(2) The popular initiative 'Against Mass Immigration' calls for the termination of the bilateral agreements with the European Union (EU).
(3) The termination of the bilateral agreements with the European Union (EU) is to be refused.
(4) The popular initiative 'Against Mass Immigration' is to be refused. 5 In its most simple form, it looks as follows (Walton 2007, p. 233 Evaluating future states of affairs in relation to the current one avoids this problem. Then, although the state of affairs with paternity leave would still be unjust as a whole, it would be recognised as more just than the current state of affairs. This would result in a positive evaluation and thus in retaining commitment to introducing paternity leave.
Macagno and Walton's scheme leaves the process of evaluating states of affairs opaque and open to interpretation. But, as we have just argued, it is crucial that one evaluates a change between states of affairs and not an isolated state. Thus, in the scheme we propose, the change will take centre stage. This will not only clarify the process of evaluation but also make it less opaque.

Reasons for Voting "Yes" or "No"
Much of what we have argued above depends on our understanding of reasons for voting "yes" or "no". In essence, we understand such a reason for someone as a change in circumstances that affects some of their concerns. This gives us a two-part structure of a reason-a change and a concern and we will discuss them in turn.

Comparing Worlds
Looking at the decision a voter is faced with, we can understand it as a choice between two worlds (alternatively, future states of affairs or circumstances): One in which the proposed state action has been accepted and enacted (the "yes"world), and one where it has been rejected (the "no"-world). 7 The question of voting "yes" or "no" then boils down to which world one prefers. Thus, we can frame a reason for voting "yes" as a reason to prefer the "yes"-world over the "no"-world (and the other way around for reasons for voting "no").
The task of the argumentative texts in the Explanations then is to make one of these worlds look appealing and the other one unappealing. For this purpose, the texts explain what will change in each case and then compare the results, though much of this is implicit.
Quite often, they compare the two worlds and frame the differences as positive or negative, depending on their side of the debate. At other times, they only describe one of the worlds and cite a positive or negative change compared to the current state of affairs-the world as it is before the vote.
This raises the question of which worlds one actually has to compare in a vote: the "yes"-world with the "no"-world, or the current one with the "yes"-world? It is worth noting that this is not a trick question, as the current world and the "no"-world are importantly different from each other (even though no specific state action has been taken in the "no"world). Two reasons: First, the world is going to change over time, and thus the world after the vote will be different from the one before it. For instance, this can be important for initiatives supposed to reduce the number of refugees coming to Switzerland. If it is to be expected that the number is going to go down anyway (e.g., because of a war coming to an end), the concerns that led to the initiative will be satisfied not only in the "yes"-world but also in the "no"-world, while not in the current world. Second, rejecting a proposed state action can have direct consequences in positively reaffirming the status quo. For instance, rejecting legislation prohibiting hate speech [see the vote on prohibiting hate speech related to sexual orientation, Schweizerische Bundeskanzlei (2020b)] can send a strong signal that hate speech is acceptable and thus lead to an increase in hate speech or discrimination against minorities. Thus, it is important to know whether one compares the "yes"-world to the "no"-world or to the current one.
Now to the question of which worlds to compare: In a sense, one has to go through both comparisons, but not quite in the same way. It seems clear that the final decision has to be based on a comparison between the "yes"-and the "no"-world. After all, one can only choose between these two worlds-it is not possible to remain in the current one. We suggest that this comparison consists in weighing the reasons for voting "yes" and the ones for voting "no". However, the individual reasons will express comparisons of the current world with either the "yes"-or the "no"-world.
One indication for why one has to take the current world into account can be found in behavioural economics. Research around prospect theory has produced considerable empirical evidence that we are loss averse (Kahneman and Tversky 1979;Kahneman 2011). That is, we give losses more weight than gains. But, in order to recognise a future state of affairs as a loss or a gain, one needs to take the current situation into account. A direct comparison between the "yes"-and "no"-worlds cannot make this distinction. Now we can state that an individual reason for voting "yes" is either a positive change in case of a "yes" or a negative one in case of a "no" relative to the current circumstances (and the other way around for reasons for voting "no"). The question that remains now is how we can characterise a positive or negative change. We suggest that it is our concerns that frame some change as positive or negative. This will be explained in the next section.

Concerns
We claimed that reasons for voting "yes" or "no" are to be found in the changes in the world that occur if the proposed state action is accepted or rejected. But, of course, not every change generates a reason-only the changes that one cares about. Where a change in the number of immigrants might well generate a reason in favour of anti-immigration initiatives, the mere change that the length of the constitution will be increased by adding an article will most likely not factor in many people's deliberation. 8 What makes one care about some change? We suggest that the answer lies in what we can call concerns (Blackburn 1998). One cares about a change if the change affects a concern. We are quite liberal with what counts as a concern, and in a more or less tautological characterisation, we might say that a concern can be anything that concerns us. This serves to highlight the subjective nature of concerns. There is no objective feature in a change that designates it as affecting a concern. For instance, if someone really cares about the number of green chairs in a city, they will have a reason to vote in favour of an initiative with the aim of increasing that number (though most likely no one else will). In our analysis, we would say that the "yes"-world presents a change (increase) in the number of green chairs and this change affects a concern of this person, thus resulting in a reason for them to vote "yes".
In the literature, one will often find that values are thought to make one care about the state of the world [cf. Macagno and Walton (2014), Atkinson et al. (2005), Fairclough and Fairclough (2012)]. That is, one might favour a proposal to reduce the number of immigrants because one thinks this will make the country safer. Here, it is the value of security that makes one care about a change in circumstances. Similarly for other fundamental values such as liberty, justice, and so on. In our terms, this just means that security and other such values are concerns. In that sense, we use "concern" as an umbrella term for anything that makes one care about some change (including values as well as green chairs).
However, in relation to values, it is important to note that, say, a projected increase in security does not always result in a reason in favour of a proposal. Namely, it only results in a reason if one actually cares about this increase, that is, one is actually concerned with increasing security. This again highlights the subjective nature of concerns. Something (a value, an interest, green chairs,...) only counts as a concern of someone if they actually care about it. Only then can they be used to positively or negatively evaluate a change.
In the claim that subjective and even idiosyncratic concerns (as with the green chairs) can generate reasons for voting "yes" or "no", we somewhat depart from the literature, at least in focus. Usually, it is assumed that reasons put forth in a public debate should make reference to some common good and not the special interest of certain groups of voters (Kock 2017, p. 285). But, while this is mostly true for the arguments put forth in the Explanations, in reconstruction debates, we should not preclude the existence of reasons based on special interest.
Generally speaking we can say that there is one crucial difference between the things that we are concerned with and the things we are not concerned with. If we are concerned with something, then changes with respect to it become potentially relevant to us. Whereas any changes remain irrelevant if we are not concerned with it.
This does not, however, mean that any change of something we are concerned with is necessarily relevant-nor does it mean that one specific change has the same effect on everyone concerned with it. Alice, Bertrand, and Chris might, for example, all be concerned with green chairs. When they see that a proposed change in the constitution has an effect on green chairs, they are accordingly all interested in what these changes are. Whereas Alice and Bertrand are mainly concerned with the number of chairs, Chris is exclusively concerned with the placement of them. Upon realising that the proposed change of the constitution merely increases the number of chairs, but does not affect their placement, this change becomes irrelevant for Chris. The change is, however, relevant for Alice and Bertrand. Since Alice likes green chairs, she welcomes this change. Bertrand, who fell from a green chair as a child, however does not. This means that even if all three are concerned with green chairs, the effect of this concern can be different. The same change is a reason to vote "yes" for Alice, to vote "no" for Bertrand and is completely irrelevant for Chris.
A proposed reason for voting "yes" or "no", then, is a change in circumstances-a change in the world-produced by either accepting or rejecting the proposed state action, that affects some concern in some specific way. Whether it is a reason in favour of or against voting "yes" or "no" then depends on whether this change is perceived as positive or negative. The subjective nature of concerns thus makes any reason person-relative. It is only for the sake of simplicity that we tend to speak about reasons for voting "yes" or "no" per se and not always make their relative nature explicit. Every reason is only a reason for someone. 9 For another example, consider the proposed reason for introducing paternity leave that cites the attractiveness of small and medium-sized businesses (SMBs) in its favour. The claim is that paternity leave will increase the attractiveness of SMBs to employees, as they, in contrast to big corporations, often cannot afford to offer paternity leave without public funding. If one is concerned with the attractiveness of SMBs in the sense that one wishes them to be more attractive, the change aligns with the personal prefences, thus making it a reason in favour of voting "yes". If, however, someone would want less attractive SMBs, say because big corporations are thought to be more beneficial to the economy, the same change would result in a reason in favour of voting "no".
These, then, are the reasons we focus on in reconstructing the debate depicted in the Explanations. Next we turn to the argument scheme for proposing such reasons. As we will see, the scheme results more or less directly from our characterisation of reasons for voting "yes" or "no".

Arguments that Propose Reasons
This section presents a scheme that captures the reasons for voting "yes" or "no" such as the ones presented above. After presenting the scheme we show how it can be used to reconstruct large parts of the argumentative texts of the Explanations, including arguments that exhibit a structure other than our scheme. For this purpose, we propose a three-layered structure of these argumentations in Sect. 6.

The Scheme for Proposing Reasons
Considering our characterisation of reasons for voting "yes" or "no", there are two main questions to ask when presented with a proposed state action: What will change in the world through accepting/rejecting the proposed state action? And which concerns do the changes affect?
With these questions in mind, we can understand the arguments put forth in a public debate surrounding a vote as suggested answers to the two questions. This gives us the following concise characterisation: A single argument picks out a specific change and a single concern, combining them to a reason. This is also a point about argument individuation. For each change and each concern, there will be an individual argument. In that way, the arguments represent proposals on what the voter should consider. Arguments put into a scheme that highlights the above structure present all basic information needed for the voter to judge whether they accept the proposed reason. Figure 1 gives the general form of the scheme, Fig. 2 contains a first example.
The core of the scheme consists of the move from projected change and concern to proposed reason. This follows from the structure of reasons described above. Once we know how our circumstances will change through the proposed state action, and we know what concern this change affects, we have a putative reason for voting "yes" or "no". That is, the scheme closes with the conclusion proposed reason that combines concern and projected change into an answer to the two questions mentioned at the beginning of this section: What will change in the world? And which Current Circumstances Currently, we have circumstances C.
Projected Outcome (Part n of) The action will achieve outcome O (in circumstances C).

Projected Change
Through (part n of) the action, circumstances will change from C to O.

Concern
Realisation of concern V is affected by the projected change from C to O in some specified way.

Proposed Reason
Through (part n of) the action, realisation of concern V will change in some specified way. Fig. 1 The argument scheme for proposing reasons

Current Circumstances
In contrast to large corporations, small and mediumsized businesses (SMBs) cannot afford to offer paternity leave.

Projected Outcome
Through federal income compensation, all businesses can afford to offer paternity leave.

Projected Change
Through federal income compensation, not only large corporations, but also SMBs can afford to offer paternity leave.

Concern
If SMBs can afford to offer paternity leave, they become more attractive to employees.

Proposed Reason
Through federal income compensation, SMBs will become more attractive to employees. concerns do the changes affect? In essence, then, such a reason represents a projected change that is evaluated in light of some concern. It is worth going through the scheme premise by premise:

Current Circumstances and Projected Outcome
The scheme starts out with two premises establishing a claim about a change the proposed state action is supposed to provoke. For that purpose, current circumstances describes some specific aspect of the state of affairs before the vote. Projected outcome then makes the conditional claim that the proposed state action is a vehicle to produce a certain outcome given certain circumstances (though these circumstances are often left implicit). In principle, this conditional claim is independent of the actual current circumstances, as it simply expresses a relation between two states of affairs. As an example consider that introducing paternity leave of 2 weeks can be an extension if it was shorter before, but a shortening if it used to be longer.
Note that "the action" in projected outcome can refer to either accepting or rejecting the state action. Both are considered "actions" for our purposes. As changes of law or the constitution can be quite complex actions where various articles are changed, deleted, or added, it can be helpful to specify which part of the action is responsible for the projected outcome. The specification "Part n of" in projected outcome allows to capture this. Such specification are almost always left implicit in the actual arguments, which can make it more difficult to determine why the outcome should occur.
The descriptions given in these two premises should be formulated in a neutral way, in essence making them accessible to fact-checking-either currently or for the outcome in the future. It is of course sometimes difficult, and some would claim impossible, to describe in a neutral way. However, aiming for neutral descriptions helps to avoid reproducing uninformative or misleading buzzwords and abstract claims.

Projected Change
Where projected outcome makes a conditional claim between some circumstances and an outcome, projected change adds the information from current circumstances to make a claim about an actual change. That is, projected change makes a claim about how the actual circumstances will change, and through that it becomes relevant for a vote.
Consider the example in Fig. 2. Looking at the outcome alone, several changes are possible. The only fixed part seems to be that currently (at the time), not all businesses can afford to offer paternity leave. However, for the concern about attractive SMBs to be affected, it is crucial that currently (at the time) SMBs cannot afford it. If, for example, currently (at the time) SMBs would offer it while large corporations did not, the concern about SMBs would not have been affected. In its place, something regarding large corporations would have had to be introduced. Thus, it is important that both projected outcome and current circumstances are considered so that we can be concerned with the right change.

Concern
Concern complements the descriptively formulated change with the proposed ground why the change should be of interest. Contrary to the change, concern is less amenable to fact-checking. Answering the question of whether we are concerned with the described change is subjective. This means that two people may agree that the projected change will happen, but disagree with respect to the concern. Since the concern is integrated in the conclusion of the scheme, proposed reason, this means that two people may agree about the projected change, but disagree about the conclusion of the scheme. Dissociating concern and projected change comes in handy in order to locate more specifically potential points of disagreement. 10 The subjectivity of concerns also means that there is no uniform or standard way Current Circumstances Businesses do not have to offer more than 2 days of paternity leave.

Projected Outcome
Because of mandatory 14 days of paternity leave, SMBs have to find short term replacements.

Projected Change
Because of mandatory 14 days of paternity leave, SMBs have to find more short term replacements.

Concern
Finding short term replacements is unsustainably effortful and expensive for SMBs.

Proposed Reason
The mandatory 14 days of paternity leave will make SMBs unsustainable. of arguing for a concern premise as there is for projected change. Hence, our scheme does not include a justification for concern as it does for projected change. Figure 3 gives another example of the application of the scheme, this time by displaying a reason against the proposed paternity leave. We end this section with a general note: The scheme is formulated in a way that makes it materially valid. However, in practice the conditional claims in projected outcome and concern might not be strong enough to exclude all possible exceptions. In that sense, the scheme is often used defeasibly.

Reconstructive Scope
Considering all we have said about our schemes and our conception of reasons for voting "yes" or "no", a natural question might be: How many of the arguments put forth in a public debate surrounding a vote can we reconstruct faithfully using the scheme for proposing reasons? We discuss two worries here: that the focus on changes means that we can only recognise consequentialist reasons, and that our scheme accounts for actions only in terms of what outcomes they are sufficient for. After all, projected outcome states that accepting or rejecting the proposed state action will achieve some outcome. But, while some actions are sufficient to produce a certain outcome, others might be necessary, though not sufficient on their own, to achieve something. Arguments based on such outcomes might seem to escape our scheme.
Alleviating the two worries will require some reframing of the arguments. Thus, the reconstructions might deviate from the original form of the arguments to some extent. But this loss in accuracy, if you will, is offset by a gain in clarity of the argumentative structure and it allows for a uniform way of reconstructing arguments.

Consequentialising
Our account of reasons focuses on changes brought about by accepting or rejecting the proposed state action. Thus, one might think that we cannot account for reasons for voting "yes" or "no" based on deontic notions such as duties. For instance, consider a proposal designed to reduce the number of refugees Switzerland is taking in. One reason to vote against such a proposal could be that one considers it our duty to take in refugees. And that reason is not bound to a specific outcome or change in future states of affairs.
The first thing to note here is the ubiquity of consequencebased arguments in political discourse. As already mentioned, Schröter finds above 90% of all direct arguments to fit the scheme for pragmatic argumentation, which is based on the effects of a proposal (Schröter 2022a, p. 61).
But, of course, the worry remains. We suggest to deal with it by what we can call "consequentialising" (Portmore 2009; Jenkins and Purves 2016) reasons for voting "yes" or "no". This means we take any reason that is not already framed in a consequentialist way and reformulate it in terms of a change in circumstances in relation to a concern (and thus consequentialise it). Using the refugee example, we can rephrase the reason based on duty to: Change: If the proposal is accepted, Switzerland will take in fewer refugees, Concern: which means that the country will be in violation of its duty to take in refugees.
It is important to note that consequentialising does not entail any substantive account of reasons. It is simply a mode of reconstructing reasons. We do not claim that consequences are all that matters for a decision or that deontic or other accounts of reasons are wrong. We only claim that it is possible to rephrase a given reason in terms of a change and a concern, nothing more.
What we have, then, is a uniform presentation of reasons for voting "yes" or "no" that we can use to cover all kinds of reasons put forth in the arguments of a public debate surrounding a vote.

On Necessary and Sufficient Actions
Practical arguments, working with reasons for actions, come in two flavours: Oftentimes, some action is sufficient to achieve some outcome, othertimes it is necessary (Macagno and Walton 2018). As the proponents of paternity leave claim, for instance, introducing paternity leave might be sufficient (though possibly not necessary) for making SMBs more attractive. According to the opponents, rejecting paternity leave might be necessary (though possibly not sufficient) to save SMBs from their doom (Schweizerische Bundeskanzlei 2020a). However, the way we presented the scheme for proposing reasons, the projected outcome premise "The action will achieve outcome O", implies that the action will be sufficient for some outcome. Thus, one might worry that we can only use our scheme for one kind of practical argument, while the other escapes it.
Here, again, rephrasing alleviates the worry. The reason for this is the intertranslatability of necessity-and sufficiency-based outcomes. Consider the following two sentences: (1) Doing action A is sufficient to achieve outcome O.
(2) Doing action A is necessary to achieve outcome O.
Our scheme can readily incorporate statements of type (1). The claim is then that the scheme also works for claims of type (2). Note first that both types of statements can be reformulated without changing their meaning. Take statements of type (2). By using contraposition, we can express the same statement as "Not doing action A is sufficient to avoid outcome O." 11 Statements of type (1) can also equivalently be expressed by their contrapositive. This means that we have the following four expressions of the two types of statements, where (1.1) and (1.2), as well as (2.1) and (2.2) are equivalent.
(1.1) Doing action A is sufficient to achieve outcome O.
(1.2) Not doing action A is necessary to avoid outcome O.
(2.1) Doing action A is necessary to achieve outcome O.
(2.2) Not doing action A is sufficient to avoid outcome O.
In the context of voting, we have two options when it comes to actions: either we accept the proposed state action or we reject it. This binary structure, reflected by the voting options "yes" and "no", 12 means that our scheme can work with statements of type (2). Take for example the claim that rejecting paternity leave is necessary to save SMBs. This claim is expressed as a statement of type (2.1), where save the SMBs is the outcome O and rejecting paternity leave is doing action A. Owing to the binary character of the choice, accepting paternity leave is the same as not rejecting it. This in turn means that accepting paternity leave is the same as not doing action A. We can thus express this claim using the structure of (2.2), namely that accepting paternity leave is sufficient for SMBs to be doomed. An analogous translation is possible if doing action A means accepting a change of legislature instead of rejecting it.
Summing up, this means that with our scheme we deal with statements of type (1.1) and of type (2.2). With this we can express both necessity-and sufficency-based reasons, either directly or by the above mentioned equivalent translation.

Reconstructing Debates Surrounding Proposed Reasons
The scheme presented above constitutes our core proposal for reconstructing the reasons and their justifications in the argumentative texts of the Explanations. However, there is more to public debates surrounding votes than is captured by the scheme. Even before going into the debate about the respective weight of the proposed reasons, more work is to be done. Namely, there is a range of arguments that matter directly for the proposed reasons but have not been discussed yet. We distinguish these arguments into two classes: First, as with every argument, the premises as well as the cogency of the arguments captured by our scheme can be debated. That is, all claims about current circumstances, projected outcomes, and affection of concerns can be supported, attacked, defended against attacks, and so on [see Toulmin (2003); for a contemporary overview see Leal and Marraud (2022)]. The second class contains the arguments that address the overall consistency and coherence of the proposed reasons as well as their structure and order.
The first class is concerned with whether the individual arguments that propose reasons hold up to scrutiny. After all, if there are fatal flaws in these arguments, the reasons they are supposed to support do not need to be considered further. This can also be readily seen in practice. For example, the debate on introducing measures to reduce Switzerland's CO 2 impact focused mostly on the effectiveness of the proposed measures [thus debating the truth of projected outcomes rather than concerns, see Schweizerische Bundeskanzlei (2021)]. Since premises such as projected outcome often use defeasible generalisations, this class of arguments can also include undercuts that do not directly attack a specific premise but an inference step. Note further that in this class, we collect the arguments discussing the reasons and their justification in isolation with little cross-reference to other reasons.
The second class then deals with the output of our core arguments, namely the reasons proposed in the argumentative texts themselves. In the end, these reasons have to come together to inform an all-things-considered claim that either favours voting "yes" or "no". Clearly, this involves somehow adding up the reasons in favour and the ones against the proposed state action and comparing their respective cumulative weight. But it seems reasonable that this process requires not just a list of isolated reasons (as produced by the scheme for proposing reasons), but a holistic picture of all reasons coming together (Dancy 2004). This is to avoid common failures of reason additivity due to hierarchies amongst the reasons or the occurrence of modifiers that are not obvious to see while the reasons are isolated [see Nair (2016)]. Hence, one has to put the individual reasons given by the texts into one coherent picture of what will happen after the vote and structure them in a suitable way.
One possible obstacle for a coherent set of reasons are incompatible reasons. Take for instance the two reasons presented in Figs. 2 and 3: "Through federal income compensation, SMBs will become more attractive to employees" and "The mandatory 14 days of paternity leave will make SMBs unsustainable." Considered in isolation, both reasons might seem acceptable. The sustainablilty and attractiveness of the SMBs both look like worthy candidates to be taken into account when deciding on how to vote. Taken together, however, there is undue friction. After all, if SMBs cease to exist, their attractiveness can hardly matter. One natural way to see the conflict between these reasons and to resolve it is that the reason about unsustainability rebuts the reason about attractiveness. From a holistic perspective, this conflict can also be described using the terminology of enablers and disablers of reasons [see Dancy (2004); Kernohan (2022)]. In such a description, the unsustainability of the SMBs leads to a change in circumstances, namely the lack of their existence, that disable the attractiveness of the SMBs as a reason in favour of voting "yes".
Another possibility is that this conflict indicates that the reasons have to be formulated more precisely. The conflict could also be resolved by reformulating or combining the reasons as: "Through federal income compensation, SMBs will become more attractive to employees for a short time, before ceasing to exist due to being unsustainable." Since the argumentation on how to consolidate such conflicting reasons is often left implicit, there is some leeway for the reconstructor on how exactly to reconstruct the arguments. In Fig. 4, we have reconstructed it as one reason rebutting the other.
Yet another important preparatory action for weighing is structuring and grouping the reasons according to their topic-dimension and level of generality. Reconstructing the argumentative texts in such a way simplifies the use of the reconstruction with respect to questions of additivity and possible problems of double counting when weighing the reasons [cf. e.g. Kearns (2016), Maguire and Snedegar (2021)]. For instance, as Kock notes, reasons in practical reasoning belong to different dimensions, that is, topics such as economy or ethics (Kock 2007). It then makes sense to compare the reasons first with other reasons of the same dimension. Also, if one reason feeds into a more general reason, they should not be added up independently of each other. For instance the reason of the unsustainability of SMBs feeds into the more general reason of the claimed damage paternity leave does to the economy. Framed differently, we can understand the specific reason of SMBs becoming unsustainable as being, amongst other things, an amplifier of the more general reason about the economy. These reasons, then, should not be added up independently.
The distinctions of classes of arguments suggest splitting up the argumentation put forth in the argumentative texts into three layers: In the middle, there are the arguments that propose reasons and are captured by our scheme. They represent the central part of the argumentation and usually take up the bulk of the texts. Then, in the upper layer, the premises and inferences of these arguments are discussed, so as to establish which reasons hold up to scrutiny and should therefore be considered when weighing the pros and cons. The last step before the weighing forms the consolidation layer, where the remaining arguments and reasons are collected and put into a cohesive form. See Fig. 4 for an example reconstruction of the three layers, though the borders between the layers may not always be as clear-cut as the figure might imply.
The result has certain similarities with what you get from argument mapping procedures such as the one proposed by Betz (2010) and Brun and Betz (2016). While their argument mapping approach is intended as a general purpose procedure, to be applied to all kinds of argumentation, the more limited scope here allows for some advantages. The use of argument schemes results in a more uniform reconstruction and representation of the debate. The arrangement of the arguments in three layers further facilitates the readability of the map.
This concludes our treatment of the reconstruction of the arguments dealing with proposing reasons in public debates surrounding votes. In terms of the whole of such debates, the main part that is missing from our picture is the part where (A1.1) The costs of paternity leave make up a much smaller percentage of the budget of big corporations than of SMBs.
(A2.1) As the absences due to paternity leave last for only two weeks, finding replacements is sustainable.
(A1) Through federal income compensation, SMBs will become more attractive to employees.

(A2)
The mandatory 14 days of paternity leave will make SMBs unsustainable.
(A3) If SMBs become unsustainable, their attractiveness becomes irrelevant. Therefore the reason from unsustainability trumps the reason from attractiveness of SMB (if both reasons are accepted).

Layer 2 Proposing Reasons
Layer 3 Consolidating Reasons Support Attack Fig. 4 The Arguments A1 and A2 are the two arguments presented in Figs. 2 and 3 (for purposes of readability, we only mention their conclusions). They are located in the middle layer. The arguments in the upper layer, A1.1 and A2.1, both discuss premises of the central arguments. A1.1 for example supports the premise Current Circumstances of A1, according to which only big corporations can afford to offer paternity leave at the moment; while A2.1 attacks the Concern of A2. Finally, in the lower layer, we have arguments that address how to consolidate reasons. A3 for instance argues that the conclusion of A2 rebuts that of A1 reasons are weighed against each other in order to come to a general conclusion on how to vote. The reasoning at this point can be broadly characterised as conductive reasoning [Wellman (1971), see also Blair and Johnson (2011)], where pro-and con-considerations are weighed to reach a conclusion of the form "All things considered, the pros/cons outweigh the cons/pros." However, while we do not discuss this part of the debate in this paper, the result from our reconstruction is wellsuited to form a basis for the weighing process. Consolidating the reasons puts them into a cohesive form that allows better weight allocation than simple disconnected lists of reasons in form of pro-con lists. There are well-documented drawbacks of such pro-con lists, such as the disregard of relations between the entries (Brun and Betz 2016;Juthe 2019), and a map like ours allows for a more fine-grained approach to voter information.
In the next section, we will switch perspectives and discuss how such a reconstruction can be put to use by a voter trying to extract reasons from the public debate surrounding a vote.

Using the Reconstruction
From the perspective of the voter confronted with a debate and a reconstruction according to the previous section, the situation can be characterised using terminology from the debate surrounding normative reasons [see Kagan (1989), Marraud (2020)]: Through the arguments in the second layer, the voter is presented with a set of proposed reasons for voting "yes" or "no". By means of the first layer, the voter then has to work through the arguments and determine which reasons need to be considered further. At the latest by now, the voter has a set of prima facie reasons. These then are the reasons that need to be consolidated into a set of pro tanto reasons. This can be done according to the third layer. Finally, they have to weigh the reasons in order to arrive at an all-things-considered reason that either favours voting "yes" or "no".
The scheme for proposing reasons enables us to structure the first part of this process along a set of critical questions. These questions serve to evaluate the strength of an argument and are formulated on the level of the scheme in order to have a unified approach for each argument fitting the scheme [see Walton et al. (2008)]. The list is not supposed to be complete [see Yu and Zenker (2020)], rather it gives an idea of the most important questions to ask when confronted with an argument of the given scheme. If all these questions can be answered, one should have a good idea of whether the argument holds up and actually justifies its conclusion.
The list of critical questions for our scheme is as follows: (CQ1) Is the description of the current circumstances true and fair? (CQ2) How likely is it that the projected outcome will occur? (CQ3) Is the concern actually affected in the way claimed? (CQ4) Do I share the concern?
With respect to (CQ1): It is important to check the factual accuracy of the claims about the current circumstances. Equally important is it to check whether a full picture is presented, "meaning that no important information and numbers are left out so that we're misled about how things really stand" (Kock 2017, p. 282). This latter aspect is reflected in the question about the fairness of the description.
When discussing any projected outcome of an action, it is of course always crucial to be aware of the fact that any projection into the future comes with some uncertainty. While the scheme itself does not highlight that, (CQ2) asks explicitly about the likelihood of projected outcomes.
(CQ3) directly asks to check the truth of the concernpremise. However, a related question that is important is whether other concerns than the one mentioned are affected. Perhaps you care about the change mentioned in the argument for some reason other than the concern suggested. Then the change should nevertheless be considered and you would end up with a reason for voting "yes" or "no" in spite of negatively answering (CQ3).
Another function of (CQ3) is to protect against certain forms of populism and hate speech, as populist arguments often exaggerate or falsely attribute affected concerns by some orders of magnitude.
Question (CQ1)-(CQ3) are part of the process of judging whether the argument actually works to propose the reason of its conclusion. But this does not mean that the voter has a reason. We understand concerns as something subjective. Thus, this last step depends on whether the voter actually shares the concern at hand. Only if this can be answered positively (CQ4), can we say that the voter has a reason. If, for instance, an argument suggests the number of green chairs as a reason for voting "yes" or "no", we would not care and not consider it as a reason to be weighed.
At the end of this process, one has a consolidated set of pro tanto reasons that are personally relevant for the decision. Of the proposed reasons, some may fail because they have not been properly established through the argumentation, others may appeal to a concern not deemed relevant. Importantly, the reconstruction need not remain unchanged throughout this process. For instance, if the voter has arguments of their own, they should be added. Additionally, arguments that do not hold up to the voter's scrutiny, can be removed. In the end, each voter should arrive at their own version, their own account of the arguments.
Once a voter has arrived at a consolidated set of pro tanto reasons, all that is left for them to do is to weigh them and conclude whether voting "yes" or "no" is favoured, all things considered. While this is easier said than done, a discussion of this process falls outside the scope of our present aims.

Further Applications
We have developed our approach solely with reference to the political landscape in Switzerland and its direct democratic features. In wrapping up, it is important to note, that our scheme for proposing reasons can also be put to fruitful use outside that context. For that purpose, we will briefly discuss a specific account of political policy deliberation and sketch how the scheme can fit into deliberation dialogues in general.
First, consider the two-phase democratic model of Bench-Capon and Prakken (2010) and Bench-Capon et al. (2011). As the name suggests, they distinguish two phases in political deliberation: the criteria deliberation phase and the proposal deliberation phase. During the first phase, the deliberators determine which criteria are to be used to assess whether the policy proposals will realise some abstract desiderata. For example, the desiderata "prosperity" can be measured using the criteria "GNP growth" (Bench-Capon et al. 2011). Based on these criteria the second phase is used to assess which proposals are best.
In our terms, what happens here is that the deliberators use the first phase to determine which changes will count as affecting which concerns. The criteria are used to make the step from a change to a concern. For instance, a change like "The proposal will achieve a growth in GNP" leads to "The action will increase prosperity" through the prior-agreed criteria-link between GNP and prosperity.
Thus, the scheme for proposing reasons fits right in with the two-phase democratic model and opens up another perspective on what is going on. Also, while Bench-Capon et al. do have argument schemes for determining criteria and for using reasons to argue for specific proposals, they lack one for how these reasons come about. This is exactly where our scheme for proposing reasons can come in.
Second, in a similar vein, the scheme for proposing reasons fits into a more general framework for deliberation dialogues proposed by McBurney et al. (2007). In their model, a deliberation dialogue goes through multiple stages, one of which is the "consider" stage. After exchanging information and proposing actions, these proposals are commented on from various perspectives, that is, they are considered (McBurney et al. 2007, p. 100).
According to the example provided by McBurney et al., this stage is used to compare the various proposals in the sense of "Participant P 1 asserts that from the perspective of the degree of risk, prohibiting the sale of phones is the lowest risk action-option possible" (McBurney et al. 2007, p. 108). If we consider the reasoning that goes into such a assertion, we see that it involves assessing how each proposal fares with regard to the concern of risk, based on a projection of the change it brings with it. Thus, here again our scheme for proposing reasons can be put to use.
Importantly, in both these cases, the context of application is much wider than Swiss voting debates.

Conclusion
We have presented an argument scheme that captures a prevalent type of reasons proposed in debates surrounding public votes on initiatives and referenda. This scheme for proposing reasons allows for a uniform presentation of these types of reasons and simplifies critical engagement with such reasons. It assists the voter in answering two important questions when voting: What will change in the world depending on how the vote goes? And which concerns do these changes affect?
Furthermore, the scheme lends itself to an organisation of the debate. It does so by distinguishing three classes of arguments. First we have the arguments proposing reasons for voting "yes" or "no". These arguments are captured by our scheme. Another class of arguments is concerned with whether certain projected changes will actually happen and what importance that they have. These arguments can be organised as reasoning about the premises and the cogency of our scheme. The third class is concerned with how the reasons come together to a consolidated picture so that they can be weighed. This class of arguments can be characterised as dealing with the conclusions of the scheme for proposing reasons.
In terms of practical applications, our approach can be incorporated in various forms of voter information. For instance, a reconstruction in our sense could be distributed as supplementary material to the Explanations, either on paper or in an interactive digital format. Importantly, our reconstructions do not provide any direct voting recommendation. Rather, they provide the material a voter can use to structure their individual deliberation. The voter has to go through the arguments themselves, with the critical questions in mind, to get to their own set of reasons they have to weigh. And then, of course, they have to figure out which way to vote is favoured by their reasons. In that sense, our reconstructions can assist and guide through some parts of the deliberation that goes into voting, but it is no substitute.
Fachschaft Philosophie Bern for opportunities to discuss early ideas. Special thanks go to Georg Brun, Pablo Hubacher, and Que Wu and two anonymous referees for helpful comments and discussions.

Conflict of interest
The authors declare that they have no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.