1 / 20

Artificial argument assistants for defeasible argumentation

Artificial argument assistants for defeasible argumentation. Bart Verheij Presented by: Jacob Halvorson. Goal: Develop experimental argument assistance systems. Reasons Administering and supervising the argument process Keeping track of: Issues raised Assumptions made

afric
Download Presentation

Artificial argument assistants for defeasible argumentation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Artificial argument assistants for defeasible argumentation Bart Verheij Presented by: Jacob Halvorson

  2. Goal: Develop experimental argument assistance systems • Reasons • Administering and supervising the argument process • Keeping track of: • Issues raised • Assumptions made • Reasons for and against a conclusion • Evaluating the justification status of the statements • Checking for user error regarding rules of argument

  3. Who would this help? • Single user: • Lawyer could use it in court • Analyze current arguments • Structure unpolished arguments • More than one user: • Argument mediation system • Keep track of diverging positions • Assist in the evaluation of opinions

  4. Isn’t this like an Automated Reasoning System? • Automated Reasoning • Knowledge base • Complex reasoning is done for user • Argument Assistant • Assist user in reasoning (not replace user’s reasoning) • Complexities are less problematic since reasoning can be left to user

  5. Argument Assistant and the Legal System • This paper focuses on using argument assistants to help with legal proceedings • Difficulties • Lack of a canonical theory of defeasible argumentation • New user interface must be designed • How to present arguments to the user • How users may perform argument moves • Legal • Legal rules are generally defeasible • Legal gap • Legal ambiguity

  6. Dialectical Theory Construction Theory of the case, applicable law and the consequences are progressively developed

  7. ARGUE! • Argument assistant based on the logical system CUMULA • CUMULA: Arguments (trees of reasons and conclusions) can be defeated • Defeat of arguments results from attack by other arguments (defeaters). • Types of defeat • Undefeated counterargument • Undefeated argument with conflicting conclusion • Problems • Not natural enough to represent real-life argumentation • On-screen drawing too complex

  8. ARGUMED 2.0 • Major problems fixed from ARGUE! • Argumentation theory focused undercutting exceptions • Reasons that block the connection between a reason and a conclusion • Support by reasons and undercutting exceptions could be represented simultaneously. • User interface is template-based • Buttons to click which will bring up a given template • Evaluated by ten testers • Show unexplained examples and try to reproduce argumentation samples in the system • ARGUMED based on DEFLOG

  9. ARGUMED based on DEFLOG • Mouse-sensitive argument screen • Add a statement box by double-clicking • Right mouse gives access to • Add support for statement • Attack against statement • Better argumentation theory • May attack any statement • Arrows between a reason and its (supported or attacked) conclusion are considered conditional statements

  10. ARGUE! and ARGUMED based on DEFLOG in a case of inflicting grievous bodily harm • Precedent 1 • Victim has several broken ribs, but no complications • Precedent 2 • Victim has several broken ribs with complications *Intentional infliction of grievous bodily harm is punishable by 8 years in prison.

  11. ARGUE! evaluation rules • A statement that is now set to justified or unjustified by the user, keeps its status. • A statement that now has justified support, is next justified. • A statement that now has no justified support and is attacked, is next unjustified. • A statement that now has no justified support and is not attacked, is next not evaluated.

  12. ARGUE! results User can arrange statements at will White box = justified Gray box = not justified Dotted line & Dot = connection no longer justified • User can set status of statement • Justified • Unjustified • Not evaluated

  13. ARGUE! Results (cont’d) Crossed out box = conclusion unjustified Limitations • Undercutting defeater can’t be challenged • If a statement is a reason for another can’t be challenged • Ex. Witness is unreliable

  14. ARGUMED based on DEFLOG examples • Allows support and attack of any statement

  15. ARGUMED based on DEFLOG evaluation rules • A statement is justified if and only if • It is an assumption, against which there is no defeating reason, or • It is an issue, for which there is a justifying reason. • A statement is defeated if and only if there is a defeating reason against it. • A reason is justifying if an only if the reason and the conditional underlying the corresponding supporting argument step are justified. • A reason is defeating if and only if the reason and the conditional underlying the corresponding attacking argument step are justified.

  16. ARGUMED based on DEFLOG results • ! = assumptions • ? = issue • Dark bold = justified • Line through text = defeated • Light italic = unevaluated Result is not justified since inflicting grievous bodily harm is not justified and even defeated by testimony

  17. ARGUMED based on DEFLOG resultsAdding testimonies of 10 pub visitors • By adding the testimonies of 10 pub visitors, the “grievous bodily harm” statement results in “unevaluated”

  18. ARGUMED based on DEFLOG resultsAttacking a statement • Adding the rule that several broken ribs with complications count as grievous bodily harm results in the accuser doing the crime

  19. ARGUMED based on DEFLOG screenshot

  20. Conclusion • ARGUE! • Developed first • Based on CUMULA which isn’t sufficiently natural to apply to real-life argumentation • Can’t attack any statement • User interface isn’t intuitive • ARGUMED based on DEFLOG • Can attack any statement • User interface allows for use of mouse to edit argumentative data on the screen

More Related