This is one of a collection of articles which has a direct, or indirect relevance for the development of the UDP. Blogger Ref http://www.p2pfoundation.net/Universal_Debating_Project
The recent events have revived my interest in an old idea… here are a few thoughts on it; let me know what you think…
The Game of UnderstandingAfter recent experiences of some people’s inability to participate in a rational discussion and their recalcitrance towards any attempt to help them to participate appropriately I have been thinking about a framework to enable people to participate in a rational discussion…
Consider the following flowchart, which is discussed in the invitation to a conversation from my blog back in 2008…
Can this basic approach be turned into a computer assisted forum that enables people to conduct a rational discussion?
The aim of the game is to generate, assess, refine and categorise logical statements regarding a topic, thereby building up a body of true statements that comprise models of the collective knowledge regarding the discussion topic. There are also collections of false statements that comprise anti-models. When a statement is judged to be true or false this is done relative to a model given its axioms and supporting statements. Hence a statement may be true in regards to some models but false in regards to others. In this way various hypotheses can be explored and tested or alternative coherent understandings can be mapped out.
The structure of the interface is designed to guide the participants, forcing them to take certain steps that are vital to the preservation of coherence within a rational discussion. In particular the calling out of disruptive participants as well as the detection of fallacies and other polemical tactics, along with the resolution of these problems within self-contained meta-discussions.
This allows the participants to rationally explore the space of coherent understandings of a topic without it devolving into antagonistic polemic about who is right or wrong, instead multiple understandings can be mapped out and assessed. Eventually the various models may be rated based on their explanatory power and self-consistency thus decisions of right or wrong can be made. However in many cases there is no right or wrong, there are just several coherent understandings, thus it is useful to map them out in the same space and see how they relate to each other; where they connect, where they differ and why.
At the end of the discussion, rather than just having a vast disorganised thread of verbiage, instead there is a structured archive of statements about subtopics, which are formed into models and anti-models of the group’s understandings of the primary topic.
Each model is built upon a set of axioms. If several different sets of axioms and hypotheses are to be tested there can be multiple models (and anti-models).
What are the primary entities involved? (for details see the object model)
- Discussion Topic
- Forum: primary, meta
- Definition (multiple allowed, each signified by a postfix on the associated term, e.g. information[CM] for information defined within the context of computational metaphysics)
- Statement: question, proposition (logical argument), opinion, axiom
- Statement Subtopic (multiple tags allowed if the statement addresses multiple subtopics)
- Statement status: irrelevant/relevant, incoherent/coherent, uncivil/civil, false/true
- Model and Anti-Model
Statements that are relevant, coherent and civil are then assessed as to whether they are true or false.
False statements are archived as failed arguments against the model (forming the anti-model).
True statements are archived as successful arguments for the model.
If there are multiple models then the statements are archived as true or false in relation to each model.
Statements are categorised by a structured voting system, where the quality of and confidence in the supporting and counter arguments is rated by participants. One cannot simply vote according to unsupported opinion and if participants attempt to distort things this will trigger a meta-discussion on their behaviour. When agreement cannot be reached regarding a statement it remains unclassified until its arguments for and against are further clarified. This process can be recursive to several levels of nesting. It is preferable that this process be completed before the primary discussion can continue onto other top-level statements.
Only a certain number of statements made by any one participant can remain unresolved – this is to disallow roaming arguments that never come to any agreement on any points along the way but just create noise in the forum. Whenever an obstacle is encountered it cannot just be ignored, it must be addressed. There can be multiple meta-discussions related to a participant if there is a need to challenge their disruptive behaviour.
The archives provide a record of all the statements made, the subtopics addressed, the terms used and their definitions.
The archive of true statements gradually builds up an agreed upon model that captures the collective understanding of the topic.
The archive of false statements gradually builds up a record of all the failed counter arguments against the model; this is the anti-model.
These archives can just be collections of statements or they can be semantically organised into structured knowledge bases.
There is only one primary forum (only one discussion topic) and as many self-contained meta-forums as are needed, one for each separate meta-discussion. The results of previous meta-discussions can be re-used when similar disruptive tactics are being challenged. The benefit of having separate meta-discussions is that the primary discussion can remain focused on the primary topic, whilst the messy and often emotional exchanges that tend to arise when people’s behaviour is called into question in a meta-discussion won’t corrupt the clarity and flow of the primary discourse.
Once a person has been called into a meta-discussion for the same offence a certain number of times, without showing any meaningful signs of recognition of their offence or any attempt to overcome it, they should immediately be expelled – because they have proven themselves unable to participate appropriately. In this way the interface is acting as an impartial moderator so that individual participants won’t have to perform the expulsion and thereby be left open to false accusations (such as silencing dissent), which could easily be believed by those who are ignorant of the protocols of a rational debate.
Statements can be linked to other statements, e.g. a false argument and the true arguments that showed it to be false. Or a true argument and the true arguments that showed it to be true. Or a statement regarding a subtopic and all the other statements that address the same subtopic. Or a statement and all of the logically derived statements that depend on the validity of the original statement. Etc… Each statement within a model has supporting or counter arguments, each of which also have supporting or counter arguments, etc, thus forming chains of statements that must eventually rest upon the axioms of the model. If the chains are still incomplete then the statement must remain uncategorised until the chains connect with the axioms via valid sequences of supporting or counter arguments.
Whenever a key term is used it must be associated with an existing definition or a new definition must be added, furthermore, each term is marked to distinguish it from other uses of the same term that have different meanings. Hence every key term that is used has an explicit definition so that trivial confusions over terms with different meanings can be easily avoided. Also any attempts to hide behind such confusion become obvious to everyone involved.
The activities of particular participants is tracked and everyone can see the number and proportion of types of statements. Thus if someone is consistently irrelevant, incoherent or uncivil this shows up in their participation record. One can also see which particular statements were made by who. Who contributed to which models or anti-models. Who focussed on which subtopics. The proportion of true and false statements made by each participant. Etc..
An important benefit of this approach is that it enables new entrants to the discussion to quickly get up to speed. Rather than just be faced with a vast edifice of unstructured verbiage, which they have to wade through and try to decipher (which most people don’t bother to do anyway), instead they are presented with a structured environment in which they can clearly see things such as:
- The primary topic and all currently discussed subtopics.
- The key terms, their definitions and their usages marked by postfix markers within the text of each statement.
- The range of models being explored.
- The axioms that each model rests upon.
- The set of existing true statements that support each model and how they were shown to be true.
- The set of previously tested and falsified counter arguments to each model (the anti-model) and how they were shown to be false.
- The record of inputs and behaviours of each of the participants, which indicates who the trouble makers are, who the protagonists and antagonists are in relation to each model, who is interested in which subtopics, etc.
Note: this game could, to a significant extent, be implemented manually in any existing forum space if people agree to a common set of protocols and curation methods. However problems often arise when a participant in the discussion tries to enforce these protocols and methods on other participants who repeatedly fail to apply them – then emotional tensions arise. Thus it is preferable if the interface itself enforces the protocols on every participant. An example of this enforcement is that the submit button will not become active until all key words have been linked to definitions – thus making it impossible to make a statement that uses key words in a vague and undefined manner.
Conceptual run-through of the main game play:An initial discussion space has certain predefined features:
* An undefined primary topic.
* An empty list of subtopics.
* An empty primary forum and several auxiliary forums for meta-discussions.
* Several empty archives: a trash archive, and for each model there is a model and anti-model archive.
* There is an empty dictionary of key terms, their definitions and their postfix markers.
* There is a game state indicator, which indicates at which point in the flow chart the discussion is currently at.
* There is an empty list of participants along with various scores for their number of contributions and the number of these that are irrelevant/relevant, incoherent/coherent, uncivil/civil, false/true (in regards to each model).
- The initial participants are registered.
- The primary topic is clearly defined.
- Any initial key terms are added to the dictionary and defined.
- Any initial subtopics are added to the list.
- In each model archive we first create axiomatic statements. Note, these can be added to or adjusted later if necessary and alternate models can also be created later. However it is easiest to set this up to begin with in order to avoid complex re-adjustments later.
The first move is made when a participant makes a logical statement regarding the primary topic.
The next move is when the group assesses the statement for relevance, coherence and civility.
- If not passed the next move is to begin a meta-discussion about the irrelevance, incoherence or incivility of the statement. This either leads to a reassessment or adjustment of the original statement or (if required) the deletion of the statement and the expulsion of the participant.
- If passed the next move is to assess the truth value of the statement in regards to each model. This may generate other statements that must then be assessed before the original statement can be assessed. This may lead to other statements, and so on. In this way statements are recursively generated and eventually connect with the axioms of the models (or new axioms are created and even new models if the axioms don’t fit any existing models). Then the process returns (back through the recursion levels) eventually resulting in the assessment of the original statement.
When a statement is assessed to be true in relation to a model:
- If in support of a model it is added to that model.
- If against a model the model must be adjusted and the ramifications of the adjustment followed through.
- If in support of a model the model must be adjusted and the ramifications of the adjustment followed through.
- If against a model it is added to that model’s anti-model.
As this proceeds a score is kept for each participant, showing the quality of their participation (which is taken into account during meta-discussions, especially in regards to the issue of expulsion). The subtopics are tracked as well, so that one