Issuepedia:Structured Debate

Overview

Issuepedia is attempting to develop a set of rules for structured debate, eventually to be turned into an open-source internet application with a web interface.

Rules

The visual structure is kind of like an upside-down chain of reasoning; we start with a conclusion, then argue over both supportive and contradictory evidence until it becomes clear whether the initial claim is adequately defended.

• Every argument starts with a target claim which asserts as fact the subject of the debate.
• Any claim may be answered by zero or more response claims
• Each response claim must relate to the target claim in one of the following ways:
  • Support: an argument that the parent claim is true
  • Counter: an argument that the parent claim is false
  • Informal (or "informational"): information which neither supports nor contradicts the parent claim but helps narrow or guide the discussion in some way
• Any response claim may also be viewed as a target claim, and the rules for target claims (given above) apply without modification
• Any claim may either be standalone or bundled:
  • A standalone claim remains active as long as it has at least one active supporting claim
  • A bundled claim remains active only as long as all of its supporting claims remain active
• Any claim which has no active counterclaims is described as "active" and possibly true
• Any claim which has at least one active counterclaim is described as "inactive" and presumed to be false
• A response claim may require a chain of reasoning in support; in this case:
  • each link of the chain becomes its own claim, subject to the same rules as any other claim
  • failure (deactivation) of any one of the links in the chain invalidates (deactivates) the claim (normal claims remain active as long as any one supporting claim remains active)

For example, a claim may have several counterclaims, but if they have all been countered as well, then the main claim remains active and will be considered true.

Refinements

• The outcome of any debate may be used as a claim, in which case the children of that debate's root claim become children of the current claim, and the same rules apply. This lets us reuse already-established truths rather than having to hash them out again each time.

In order to make InstaGov participants more accountable for their decisions, we might require that every issue up for vote be stated as a structured debate claim. If an individual votes against the conclusion reached by the debate, we might require them to sign off on each supporting argument they are thereby disagreeing with, or perhaps display each active counterargument in a publicly visible way. This will put some "peer pressure" on individuals not to simply ignore the results of the debates, require some accountability for those who do, and also give a better idea of which premises are being taken less seriously.

There should also be some way for individuals to agree or disagree with any claim in any debate. This would:

• let us get a very quick and dynamic indication of how much support there is for any given "fact" in our debate database
• let us provide, as an almost competitive kind of thing, individuals to be notified when claims they have "endorsed" are contradicted, e.g. "Your claim has been challeged! You agreed that (text of claim), but this has been countered with (text of counter). Do you wish to respond? [link]"

Some further refinements will be necessary when adapting this system for making time-dependent decisions (see InstaGov).

Potential Issues

Chewbacca participants

The one major problem which seems likely to raise its head is that of an unfriendly participant (UP) posting nonsensical arguments which the system will automatically count as valid, thereby requiring a counter. Although countering them may be just as quick as creating them (e.g. "This is nonsensical"), the argument's visual presentation could be rapidly overwhelmed by the nonsense-and-counters and become practically unreadable.

There are several possibilities for dealing with this. An obvious one, which may be the best solution, is to offer the option to vote on comment relevance; comments below a certain threshhold (which each user may set for her/himself) are automatically hidden/suppressed.

Quote mapping

Another, somewhat less thorny problem is involved in the process of "mapping" an existing freeform debate into a structured debate. Claims in freeform format are often tightly bundled together and need to be "unrolled" and disambiguated. What we need is some way to take the original quote, mark it up with the claims it seems to represent, and then insert those claims into the structure of the argument while referencing the original quote.

A semi-obvious way of dealing with this is simply to treat quotes as sources. This does open up the question, however, of how to handle authoritativeness and misrepresentation; perhaps "source" needs to be a data entity understood by the system, and sources whose claims are repeatedly contradicted need to have a lower "authority" score than sources whose claims are not, or whose claims are repeatedly confirmed by other sources. Although this makes the programming substantially more complicated, tentatively it would seem a worthwhile thing to spend significant time on (perhaps not in the first version, however).

Example

(using text terms rather than icons)

• claim: Socrates is mortal
  • support: Socrates is mortal because he is a man.
    • requirement: All men are mortal.
    • requirement: Socrates is a man.
  • support: Socrates is dead, therefore he was mortal.
    • requirement: Socrates is dead.
    • requirement: Death is sufficient to demonstrate mortality.
  • counter: Socrates's works have endured for millennia, therefore he is immortal.
    • counter: This is an argument that Socrates's works are immortal, not that he himself is immortal.

Exploratory Option

It looks like it would be useful to have an option for a less rigorous but still structured debate, where territory is still being mapped out and participants are not so much disagreeing with each other as engaging in a sort of question-and-answer volley. A good example is here. "Exploratory" seems like a good name for this mode. It would omit the tracking of pro-and-con and focus more on identifying the individual participants, which establishes individual beliefs and positions at various locations in the issue's "terrain" without necessarily invoking conflict.

Later on, we might add categorization-tagging of each point so we could (for example) quickly look up all of a given participant's statements on a given issue, or all participants' statements on that issue. (This would also require the ability for participants to go back and clarify or comment on their positions, especially if they change in the light of later evidence.)

Debaticons

In order to make the flow and status of structured debate easier to understand, Issuepedia has developed a set of icons and associated templates.