Discussion: Difference between revisions

Latest revision as of 08:46, 15 August 2021

This page is a knowledge crystal of subtype method. The page identifier is Op_en2382
Moderator:Jouni (see all)
Citation of this page: Jouni T. Tuomisto, Mikko V. Pohjola: Discussion. Introduction to discussions in Opasnet.. Opasnet 2019. [1]. Accessed 19 Nov 2024.
Upload data {{#opasnet_base_link:Op_en2382}}

Discussion is a method to organise information about a topic into a form of hierarchical thread of arguments trying to resolve whether a statement is true or not. In discussion, anyone can raise any relevant points about the topic. Discussion is organised using the pragma-dialectical argumentation theory^[1]. A discussion usually consists of three parts: 1) opening statement(s); 2) the actual discussion organised as hierarchical threads of arguments; and 3) closing statement(s), which is updated based on the discussion, notably any valid arguments pointing to it. When a closing statement is updated, the content should be accordingly portrayed within texts that refer to the discussion.

Contribution in the form of remarks or argumentative criticism on the content of wiki pages is most welcome. It can change the outcome of an assessment; it will improve it and make the assessment better understandable for decision makers and other stakeholders. The discussions will show the reasoning behind the work done in an assessment; it will indicate the objective and normative aspects in an assessment. In this way, decision makers and stakeholders can judge themselves whether they agree on such normative weightings. Discussion rules and formats facilitate the execution and synthesis of discussions.

Question

How should discussions be organised in such a way that

they can capture all kinds of written and spoken information, facts and valuations related to a specific topic,
there are straightforward rules about how the information should be handled,
the approach facilitates the convergence to a shared understanding by easily identifying and describing differing premises and other reasons behind disagreements,
the appraoch can be applied both a priori (to structure a discussion to be held) and a posteriori (to restructure a discussion already held)?

Answer

Discussion structure

How to read discussions

Fact discussion: Example discussion showing a typical structure (Disc1)

Opening statement: Opening statements about a topic. This is the starting point of a discussion.

Closing statement: Outcome of the discussion, i.e. opening statement updated by valid arguments pointing to it.

(Resolved, i.e., a closing statement has been found and updated to the main page.)

Argumentation:

⇤--arg1: . This argument attacks the statement. Arguments always point to one level up in the hierarchy. --Jouni 17:48, 8 January 2010 (UTC) (type: truth; paradigms: science: attack)

←--arg2: . This argument defends argument arg1. --Jouni 17:48, 8 January 2010 (UTC) (type: truth; paradigms: science: defense)

←--arg3: . This is an invalid defense of arg1 because it is successfully attacked by argument arg4. --Jouni 17:48, 8 January 2010 (UTC) (type: truth; paradigms: science: defense)

⇤--arg4: . This is a valid attack against argument arg3, because it is itself not successfully attacked. --Jouni 17:48, 8 January 2010 (UTC) (type: truth; paradigms: science: attack)

--→arg5: . This is a branch. The argument one level higher (arg4 in this case) defends this argument, but this argument points to a new statement, not the original one of this discussion. The use of branches is not recommended, as they don't bring added value to this question; rather, start another discussion with a relevant statement. --Jouni 10:10, 10 January 2010 (UTC) (type: truth; paradigms: science: branch)

----arg6: . This is a comment. It clarifies the discussion but does not invalidate arguments. --Jouni (talk) 17:38, 6 December 2014 (UTC) (type: truth; paradigms: science: comment)

Discussion rules

Freedom of opinion. Everyone has the right to criticize or comment on the content of a discussion.
A discussion is organised around an explicit statement or statements. The purpose of a discussion is to resolve which of the opening statements, if any, are valid. The statement(s) are updated according to the argumentation; this becomes the closing statement.
A statement is defended or attacked using arguments, which themselves also can be defended and attacked. This forms a hierarchical thread or tree-like structure.
Critique with a supporting, attacking, or commenting argument is stated in connection to what is being criticized.
Argumentation must be relevant to the issue that they target.
Only statements made and arguments given can be attacked.
An argument is valid unless it is attacked by a valid argument. Defending arguments are used to protect arguments against attacks, but if an attack is successful, it is stronger than a defense.
Attacks must be based on one of the two kinds of arguments:
- The attacked argument is claimed to be irrelevant in its context.
- The attacked argument is claimed to be not true, i.e. it is not consistent with observations.
Other attacks such as those based on evaluation of the speaker (argumentum ad hominem) are weak and are treated as comments rather than attacks.
Argumentation can not be redundant. If arguments are repeated, they should be merged into one.
You are supposed to be committed to your statements, that is:
- if someone doubts your statement or argument (comment), you must explain it (edit or defend).
- if someone attacks your statement or argument (attack), you must defend it (defend).
A discussion is called resolved, when someone writes a closing statement based on the opening statement and the current valid arguments targeting it, and updates the text (typically on a knowledge crystal page) that is targeted by the discussion.
However, discussions are continuous. This means that anyone can re-open a discussion with new arguments even if a closing statement has been written.

Rationale

The structure of the discussion follows the principles of the pragma-dialectics.^[1]R↻

A discussion is typically an important detail of a larger whole, such as a knowledge crystal in an assessment, but does attempt to give a full answer to the knowledge crystal question. The purpose of a discussion is to identify which of the opening statements are valid, or how they should be revised to become valid.

Arguments are actually statements; the only differences is that the target of an argument is another argument or statement within a particular discussion, while the target of a statement is some explicated use outside the discussion, such as in the rationale of a knowledge crystal. Therefore, an argument can be upgraded into a statement of a new discussion, if it is needed elsewhere.

How to discuss

Open policy practice embraces participation, in particular deliberative participation. Therefore all contributions in the form of remarks or argumentative criticism on the content of the assessments, variables, methods as well as other content are most welcome. The contributions can change the outcome of the assessments by improving their information content and making it better understandable for decision makers, stakeholders and public. Documented discussions also show the reasoning behind the work done in assessments making it possible for decision makers, stakeholders and public to judge for themselves whether they agree with the reasoning behind the outcomes. In order to obtain an orderly discussion, rules and format for discussion in open policy practice have been created based on pragma-dialectics, a systematic theory of argumentation.

Discussion has a central role in the collaborative process of formulating questions, developing hypotheses as answers to these questions, and improving these hypotheses through criticism and corresponding corrections. When a diverse group of contributors participate in an assessment, it is obvious that disputes may arise. Formal argumentation offers a solution also to clarify and potentially solve disputes. In collaborative assessments, every knowledge crystal and every part of them are subject to open criticism according to the rules modified from pragma-dialectics^[1]): see Answer above.

When a discussion goes on, there is often a need to clarify the opening statement to make it better reflect the actual need of the discussion. Within a small group of actively involved discussants, the statement can be changed with a mutual agreement. However, this should be done with caution to not distort the original meaning of any existing arguments. Rather, it should be considered whether a new discussion with the revised opening statement should be launched.

Discussion structure

A discussion has three parts: opening statement(s), argumentation, and closing statement(s). Often also references are added to back up arguments. These are briefly described below using a discussion template. Argumentation consists of defending and attacking arguments and comments.

How to read discussions

Fact discussion: Statments accepted except if toldya. (Disc2)

Opening statement: Opening statement contains one statement or several alternative, conflicting statements. This explicates the dispute at hand. In Opasnet it must be relevant for the page where the discussion is located.

Closing statement: Closing statement contains the current valid statement of the discussion, revised based on the opening statements and the valid arguments targeting it. In this example, the current arguments indicate that the opening statement is accepted, except if you apply paradigm toldya then it is not. The content of a closing statement is transferred to the texts that refer to this discussion (in Opasnet, such references typically come from a knowledge crystal page to its own talk page where the discussion is); after this, the discussion is called resolved. It should be noted that resolutions are always temporary, as discussions can be opened again with new arguments.

(Resolved, i.e., a closing statement has been found and updated to the main page.)

Argumentation:

----arg1: Argument structure explained The argumentation contains the actual discussion, organised as hierarchical threads of arguments. Each argument is either an attack against or a defense for an argument (called target). The original statement can also be targeted. As an argument always point to another argument, they form a hierarchical thread structure. It is also possible to use coordinative arguments where two or more arguments together act like one argument. Each argument is valid unless it has no proponents (a discussant promoting the argument) or it is attacked by a valid argument. However, also other validity rules than the default one can be used (see #validity and relevance below). In addition to attacks and defenses, also comments can be used for asking or offering clarification; comments do not affect the validity of the target argument. For example, this paragraph is a comment. --Jouni 23:57, 2 January 2010 (UTC) (type: truth; paradigms: science: comment)

←--arg2: Defense explained If you agree with the argument one level higher (the opening statement in this case), you should use this defending argument template. --Jouni 23:57, 2 January 2010 (UTC) (type: truth; paradigms: science: defense)

⇤--arg3: Attack explained If you disagree with the argument one level higher (the opening statement in this case), you should use this attacking argument template. --Jouni 23:57, 2 January 2010 (UTC) (type: truth; paradigms: science: attack)

⇤--arg933: Paradigm-specific argument This attack claims that the target argument arg3 is untrue. However, this claim is disputed, and with scientific paradigm this argument is untrue (because the scientific paradigm requires that every argument is backed up by references). However, with paradigm toldya this argument is true (because with toldya, there is no such requirement). With both paradigms, the attack against arg3 is relevant. --Jouni (talk) 06:43, 14 July 2018 (UTC) (type: truth; paradigms: science: attack, toldya: attack)

←--arg4: Invalid argument explained This argument is invalid because it is attacked by a valid argument (arg5). If you want to make it valid again, you should successfully invalidate all the attacking arguments. --Jouni 23:57, 2 January 2010 (UTC) (type: truth; paradigms: science: defense)

⇤--arg5: Truthlikeness attack This is the argument that attacks the truthlikeness of argument arg4. The id numbering does not have any specific order or meaning, they are just used for identifying arguments, but it is recommended that it starts with arg. Typical numbering systems are to use the first unused natural number a random four-digit number. Id is used as a reference when needed. --Jouni 23:57, 2 January 2010 (UTC) (type: truth; paradigms: science: attack)

⇤--arg1124: True but irrelevant argument. This argument attempts to attack the truthlikeness of argument #4, but this argument is irrelevant (and thus ineffective) although it is true. --Jouni (talk) 08:31, 12 July 2018 (UTC) (type: truth; paradigms: science: attack)

⇤--arg1125: Irrelevance attack This argument attacks the relevance of argument #1124 and says that it is irrelevant in attacking the statement. --Jouni (talk) 08:31, 12 July 2018 (UTC) (type: relevance; paradigms: science: attack)

The figure above shows a discussion started by Bengt Holmstöm about problems of open governance. Each argument is shown as a trapezoid. Discussion is organised around an opening statement (pink), which develops into a closing statement (blue for facts, green for values) during the discussion process.

Structure of an argument

Each argument has the following properties (see table below).

**The parameters of an argument and possible combinations.**
Id	Title	Content	Sign	Target	Type	Paradigm	Relation	Result	Comment
arg1234	Short title for display	Actual argument	Signature	arg9876	relevance	science	attack	1	If paradigm changes (all else equal), relation may change, although typically only the result changes.
arg1234	Short title for display	Actual argument	Signature	arg5555	relevance	science	comment	0
arg1234	Short title for display	Actual argument	Signature	arg6666	truth	science	defense	1	Truth refers to the truth of the target
arg1234	Short title for display	Actual argument	Signature	arg1234	selftruth	science	attack	0	Selftruth refers to the truth of the argument itself, unlike other types that refer to the target.
arg1234	Short title for display	Actual argument	Signature	arg9876	relevance	toldya	comment	0
arg1234	Short title for display	Actual argument	Signature	arg5555	relevance	toldya	defense	1
arg1234	Short title for display	Actual argument	Signature	arg6666	truth	toldya	attack	0
arg1234	Short title for display	Actual argument	Signature	arg1234	selftruth	toldya	comment	1	The relation in case of type=selftruth is irrelevant and is ignored.
									These are unique to an argument
									These are unique to an argument-target pair
									These are unique to a triple of argument-target-paradigm

Importantly, an argument always has the same id, title, content, and signature. Even if the argument is used several times in different parts of a discussion, it is still a single argument with no variation in these parameters. However, an argument may target several other arguments (as shown as an arrow on insight network graph). Each of these arrows has exactly one type (either relevance or truth); if an arguments targets itself, the type is selftruth.

Finally, people may disagree about the target relation (whether an argument is attacking, defending, or commenting a target argument) and also whether the target relation is successful or not. These disagreements are operationalised as paradigms. One paradigm has exactly one opinion about the relation and the result (e.g. that an argument is an untrue attack), while another paradigm may have another conclusion (e.g. that an argument is true defend).

Parameters are defined in the argument template of Opasnet, and they are embedded into the html code when a wiki page is parsed. It is therefore possible to collect that data by page scraping. The following properties are used to identify the properties of arguments.

**Parameter properties**
Parameter	Css selector (Opasnet page scraping)	Requirements
Id	.argument attr=id	Must start with a letter
Title	.argument .title	Short text. Is shown on insight graph as node label
Content	.argument .content	Text, may be long. Is shown with hover on graph
Sign	.argument .sign a:first-of-type	Must contain a link to participant's user page. Is shown with hover on graph
Target	NA	Previous argument one level up, or the statement for arguments on the first level
Type	.argument i.type	One of the three: relevance, truth, or selftruth (or "both", which is depreciated)
Paradigm	.argument .paradigm	Each paradigm should be described on a dedicated page. The rules implemented must be clear
Relation	.argument .relation	Is one of these: attack, defense, comment. "Branches" are typically uninteresting and ignored.
Result	relevance= .argument .relation attr=color. Gray= 0 (irrelevant), other=1 (relevant). truth= .argument .relation attr=color. Gray=0 (untrue), other=1 (true) selftruth= .argument .selftruth attr=color. Gray=0 (untrue), other=1 (true)	Truthlikeness of the relation. Either 1 or 0

Validity and relevance

Main article: Paradigm.

Each argument may be valid or invalid meaning that it does or does not affect its target argument, respectively. Validity depends on two parameters of an argument: it is valid if and only if it is true and relevant. Arguments that are untrue or irrelevant are invalid. It should be noted that with arguments, truth and relevance are thought in a narrow, technical sense: if an argument fulfills certain straightforward truth criteria, it is considered true, and the same applies to relevance. This is not to mean that they are true or relevant in an objective sense; rather, these are grassroot-level practical rules that ideally makes the system as a whole to converge towards what we consider truth and relevance. The current default criteria (called the "scientific paradigm") are the following.

An argument is true iff

it is backed up by a reference, and
it is not attacked by a valid argument about its truthlikeness (a truth-type argument).

An argument is relevant iff

it is not attacked by a valid argument about its relevance (a relevance-type argument).

Truth is a property of an argument itself, so if it is true in one discussion, it is true always. Of course, this does not mean that a sentence used in an argument is true in all contexts, but rather that the idea presented in a particular context is true in all discussions. Therefore, people should be very clear about the context when they borrow arguments from other discussions.

In contrast, relevance is a property of the relation between an argument and its target argument (or target statement). Again, this is a context-sensitive property, and in practice, it is possible to borrow relevance from another discussion only if both the argument and its target appear in that exact form and context in both discussions.

Each argument is an attack (red), a defense (green), or a comment (blue) towards its target. The nature of the argument is shown with its colour. The same argument may also attack or defend another argument, with possibly a different colour. This is because the colour is actually not the colour of the argument itself, but it is the colour of its relation with the target. There are a few possibilities to avoid confusion with these differing colours when using arguments on a wiki page.

An argument is written once in one place, and then a copy of it (with only the arrow and the identifier) is pasted to all other relevant places, with proper colours for those relations.
I there are several opening statements, the colour should always reflect the relationship to the first (i.e., primary) statement. If the primary statement changes, the colours should be changed respectively.

The legacy templates (Attack, Defend invalid etc.) do not differentiate between truth and relevance, but only validity. Therefore they are depreciated, and a new generic template Argument should be used instead. It is capable of showing relevance (irrelevant arguments have gray arrows) and truth (untrue arguments have gray content). If a legacy template is used, an invalid argument is assumed to be both irrelevant and untrue; it is also assumed to apply the scientific paradigm. If other paradigms are used, this must be stated clearly in the text, because the template offers no functionality for it. In Opasnet, the Argument template is capable of describing five different paradigms and the relevance and truth values of each (see #Practices in Opasnet).

Paradigms in argumentation

Paradigms are collections of rules to determine when an argument is true or relevant. The scientific paradigm is the default in Opasnet, but any paradigms can be developed as long as the rules can be explicitly described and implemented. For example, previously Opasnet implicitly applied a paradigm called unattackedstand (although the name was coined only in summer 2018 and the mere concept of paradigms was developed in early 2018). Unattackedstand has the same rules as the scientific paradigm except that a true argument does not need a reference, a user backing up an argument with their signature is enough.

Paradigms may also have other rules than direct validity rules. For example, the scientific paradigm considers an argument based on observations stronger than an argument based on (expert) opinions without observations, and an argument ad hominem is even weaker.

However, the rules in a paradigm can be anything, e.g. that the strongest arguments are those by a particular user or an authoritative source, such as a holy book. These rules will clearly lead to different validity estimates and interpretations of a discussion. But the methods of discussions and open policy practice have been developed having this in mind. The outcome of such explicitly described differences in interpretations are called shared understanding, and that is considered the main product of these methods.

Practices in Opasnet

For discussing, the #discussion structure should be used. In Opasnet, click the blue capital D in the toolbar on top of the edit window to apply the discussion template. This is how the discussion format appears:

{{discussion 
|id = unique identifier of discussion on this page
|Statements = 
|Resolution = 
|Resolved = Yes, if respective texts updated; empty otherwise.
|Argumentation = Threaded hierarchical list of arguments. Each argument is on its own line. Hierarchy is created by using indents (colon character : in the beginning of a line). For example:

{{argument|relat1=comment|id=1|content=The blue horizontal line on the toolbar represents the comment button. It yields this blue layout, which is used for comments and remarks.}} 
:{{argument|relat1=attack|id=3|content=This red arrow represents an attacking argument. }} 
::{{argument|relat1=defend|id=2|content=This green arrow represents a defending argument.}} 
}}

Arguments can have the parameters that are listed below (each parameter is shown on a separate line for clarity). Note that the parameters may be in any order, and it might be a good idea to show relat1 first. For details, see Template:Argument.

{{argument
| id = identifier of the argument, unique on this page, default: arg + 4 random digits
| title = a short description of the content; displayed on insight networks
| content = content of the argument
| sign = signature of the speaker, default: --~~~~
| type = type of the relation to the target argument, i.e. what is attacked or defended. Either truth or relevance.
| parad1 = main paradigm used to derive the relation between this argument and its target. This is used to format the argument.
| relat1 = relation type between the argument and its target according to the first paradigm: it has two words separated by a single space. The first is either relevant or irrelevant and the second either attack, defend, or comment.
| true1 = truthlikeness of the argument according to the first paragism: either true or untrue.
| parad2 = the second paradigm used.
| relat2 = the second relation type according to paradigm2. Default: relat1
| true2 = the second truthlikeness according to paradigm2. Default: true1
| parad3 etc. up to parad5 in this wiki
}}

Furthermore:

If you agree with an argument made by others, you can place your signature (click the signature button in the toolbar) after that argument.
Arguments may be edited or restructured. However, if there are signatures of other people, only minor edits are allowed without their explicit acceptance.

In order to contribute to a discussion you need to have a user account and be logged in.

Referring to a discussion in Opasnet

On a text that refers to a particular discussion (often on the content page of the respective talk page), you should make links at the relevant points to the respective discussions. There are two possibilities:

D↷ Link to a discussion that is not yet resolved.
R↻ Link to a resolved discussion.

Because all discussions can be re-opened, the difference between the two is not whether people are likely to participate in the discussion in the future or not. Instead, R↻ means that the current outcome of the discussion, whether an agreement or a continuing dispute, has been transferred to the main page, i.e. the contents of the main page reflect the current status of the discussion. In contrast, D↷ means that in the discussion itself, there is some information that is not yet reflected on the main page; therefore, the reader should read the discussion as well to be fully aware of the status of the page. This way, there is not a need to constantly update the main page during an active discussion. The updating can be done when the outcome of the discussion has stabilised.

Re-organising discussions afterwards

Free-format discussions can be re-organised a posteriori (afterwards) into the discussion structure presented here. The main tasks in this work are to

document original sources of material,
remove redundant text,
structure the arguments around a useful opening statement,
clarify arguments to be understandable without the context of the original discussion,
analyse and synthesise outcome into a closing statement,
save and publish your work.

See an example of a re-organisation work in Discussion of health effects of PM2.5 in Finland(in Finnish)

Calculations

+ Show code - Hide code

# This is code Op_en2382/update on page [[Discussion]]

library(OpasnetUtils)

update_truth <- function(
  pa, # P(A). This is updated to P(A|B).
  pb, # P(B)
  relb # parameter rel(B) for producing se = P(B|A)
) {
  if(any(pa>=1) | any(pa<=0)) stop("probability P(A) must be between ]0,1[, not ",pa)
  if(any(pb>=1) | any(pb<=0)) stop("probability P(B) must be between ]0,1[, not", pb)
  if(any(relb <= -1) | any(relb >=1)) stop("relb must be between ]-1,1[, not", relb)
  if(relb>=0) {
    se <- pb + relb * (pmin(1,pb/pa) - pb)
  } else {
    se <- pb + relb * pb
  }
  pab <- pa * se / pb
  return(pab)#list(pab=pab,se=se))
}

update_relevance <- function(
  relb, # relevance parameter prior for argument B: rel(B), ]-1,1[. This is updated.
  pci, # vector of probabilities P(C_i)
  relci # vector of relevance parameters for parent arguments C_i: rel(C_i), ]-1,1[
) {
  if(any(relb <= -1) | any(relb >=1)) stop("relb must be between ]-1,1[, not", relb)
  if(any(pci>=1) | any(pci<=0)) stop("probability P(C) must be between ]0,1[, not", pci)
  if(any(relci <= -1) | any(relci >=1)) stop("rel(C_i) must be between ]-1,1[, not", relci)
  relb_sign <- sign(relb)
  out <- abs(relb)
  relci_weighted <- relci * pci
  for(weight in rev(sort(relci_weighted))) {
    if(weight>=0) {
      out <- out + weight*(1-out)
    } else {
      out <- out + weight*out
    }
  }
  out <- out * relb_sign
  return(out)
}

infer_tree <- function(df, verbose=FALSE) {
  args_by_level <- df$Item[order(-df$level)]
  columns_inherited_from_alias <- c("text","truth")
  for (arg in args_by_level) {
    if(length(df$Item[df$Item==arg])!=1) warning("Argument ",arg," is not unique.")
    parents_truth <- df$Item[df$Object == arg & df$class=="truth"]
    parents_relevance <- df$Item[df$Object == arg & df$class=="relevance"]
    alias <- df$alias[df$Item==arg]
    if(alias!="") {
      df[df$Item==arg,columns_inherited_from_alias] <- df[df$Item==alias,columns_inherited_from_alias] 
    }
    for(parent in parents_truth) {
      pa = df$truth[df$Item == arg]
      pb = df$truth[df$Item == parent]
      relb = df$relevance[df$Item == parent]
      out <- update_truth(pa, pb, relb)
      if(verbose) print(c("truth",arg, parent, signif(c(pa,pb,relb, out),3)))
      df$truth[df$Item == arg] <- out
    }
    if(length(parents_relevance)>0) {
      relb = df$relevance[df$Item == arg]
      pci = df$truth[df$Item %in% parents_relevance]
      relci = df$relevance[df$Item %in% parents_relevance]
      out <- update_relevance(relb, pci, relci)
      if(verbose) print(c("relevance",arg, parent, signif(c(relb, pci, relci, out),3)))
      df$relevance[df$Item == arg] <- out
    }
  }
  return(df)
}

objects.store(update_truth, update_relevance, infer_tree)
cat("Functions update_truth, update_relevance, and infer_tree stored.\n")

References

↑ ^1.0 ^1.1 ^1.2 Eemeren, F.H. van, & Grootendorst, R. (2004). A systematic theory of argumentation: The pragma-dialectical approach. Cambridge: Cambridge University Press.

[pragmadial-1] 1.0 ^1.1 ^1.2 Eemeren, F.H. van, & Grootendorst, R. (2004). A systematic theory of argumentation: The pragma-dialectical approach. Cambridge: Cambridge University Press.

[1]

Discussion: Difference between revisions

Latest revision as of 08:46, 15 August 2021

Contents

Question

Answer

Discussion structure

Discussion rules

Rationale

How to discuss

Discussion structure

Structure of an argument

Validity and relevance

Practices in Opasnet

Referring to a discussion in Opasnet

Re-organising discussions afterwards

Calculations

See also

References

Navigation menu