Grasbon & Braun

Working Notes, by Zach Tomaszewski

for ICS 699, Fall 2005, directed by Dr. Kim Binsted

Main article

Grasbon, Dieter, and Norbert Braun. "A Morphological Approach to Interactive Storytelling." Digital Storytelling Department, Computer Graphics Center, Darmstadt, Germany, 2001. <http://netzspannung.org/version1/extensions/ cast01-proceedings/pdf/by_name/Grasbon.pdf >

Context

• Morphological approach the basis of Grasbon's 2001 thesis (in German). Summarized in article for netzspannung.org/journal.
• Grabson's site (grasbon.com): A couple 2002 talks/presentations based on using lucid dreaming as an approach to IN and VR. Combining IN with augmented reality to have a historical story while navigating a city.
• No new publications in past 3 years from Grasbon.
• Other IN projects at http://www.zgdv.de/zgdv/departments/z5 -- AI character, augmented reality, etc.

Main Points

• Focus on high-level control of plot; leave details of scenes to authors.
• Prototype environment is a physical augmented reality space.
• Plot elements tied together by cause and effect ["c/e relationships more important than events themselves"?], so separate modules not possible. Propp's functions can guide us here.
• "Each scene has to correspond with a function in the story model and is annotated with information about its minimum and maximum duration, context requirements, characters, setting, levels of violence, etc., unless these factors are generic."
• "Story act" is feedback from user after/for each scene. (Currently, text scenes and user chooses of possible outcomes.)
• Propp functions abstracted away from particular characters, settings, and action. Function sequence selected in real time; user feedback can affect sequence.
• Scenes are like instances of function classes.
• Polymorphic functions are those whose outcome is not yet determined at the start; it's result is determined by user. (Content of all scenes can be flexible, but not all functions are flexible in this polymorphic way.) Example: Propp's A (attempts of villainy) and E (hero tested or riddled) may have different outcomes. Can be repeated with different scenes or non-polymorphic variant when certain outcome needed.
• Functions have some constraints--"any D can follow A, but each D requires an E." Other constraints: time limit, current physical location, scence context requirements (introduce character, require certain misfortune state, resolve misfortune state), user state (age/maturity, bored/entertained/overloaded [eventually]). If multiple scenes fit requirements, random choice b/w them.
• Details available for Prolog-implementation processing of functions.
• Future: Generate scenes dynamically/on demand. [Me: yet that was strength of this approach--computers suck as scene details.] Finish implementation.

Related/Resulting Work

• Ulrike Spierling, Dieter Grasbon, Norbert Braun, Ido Iurgel. "Setting the scene: playing digital director in interactive storytelling and creation." Computers & Graphics 26 (2002) 31-44. <http://www.zgdv.de/zgdv/files/zgdv/departments/z5/Z5Publications/2002_02/document.pdf>
  • Outlines a level-based approach to IN, with multiple run-time engine handling different details, from high-level plot to individual character responses. [Me: Planning for multi-user system might help, remembering Laurel's conception of agents. Some agents that follow directions (AI) and some that don't (user)?]
  • Emergent narrative from bottom-up, or top-down approach from a narrative "genome"?
  • Interactive story-creation (user affecting creation) vs. story telling (user affecting delivery).
  • What role are we playing? Director, editor, actor, stage director, casting director, playwright?
  • Hard to clarify methods withough experimention; difficult to experiment with creation without tools; can't build tools without an idea of method.
  • Starting with agents (w/ only some direction) (Bates/Oz); or start with plot (w/ only some character driving) (most other projects); or start at very high-level functions (Propp, et al.)?
  • GEIST project [project from above, now named]--historical augmented reality about ghosts. Some plot (historical narrative) inflexible; other parts interactive.
  • 4 level model proposed: Story structure, scene details, character action, presentation. Output of one is input of next. [But no feedback?] Each level can be more or less predefined or autonomous. [Other models we've seen: (Idea), Action, Character, Thought; Plot, Characters, Setting]
  • Narratives are particular.
  • Emergent narrative provides too many extraneous details and fails to produce sufficiently structure narratives.
  • Details given for functioning of each level, with examples of different levels of predefined/autonomous behavior depending upon which of 2 applications each is applied to.
• Sean M. Falconer, David Gay, Lev Goldfarb's "ETS representation of fairy tales" (etc), 2005.
  • Falconer's CS Master's thesis; Goldfarb as chair.
  • Evolving Transformation System. Event-based representations.
  • Focused on representing Russian folktales for purposes of IR. Diagramming system, and suggests that perhaps we inductively learn general folktale structures.
• Chris R. Fairclough and Pádraig Cunningham's "AI Structuralist Storytelling In Computer Games" and "A Multiplayer Case Based Story Engine".
  • Irish duo pulling from all the projects we're looking at. [Follow up on this separately.]
• Fred Charles, Miguel Lozano, Steven J. Mead, A.F. Bisquerra, and Marc Cavazza's "Planning Formalisms and Authoring in Interactive Storytelling."
  • Explores tradeoffs between using Hierarchial Task Network and Heuristic Search Planning for story planning. HTN means stronger narrative coherence; HSP tends to be more flexible. Explores IN tension between character-based and plot-driven approaches.
• Belén Díaz-Agudo, Pablo Gervás, and Federico Peinado's "A Case Based Reasoning Approach to Story Plot Generation".

  We propose a Knowledge Intensive CBR (KI-CBR) approach to the problem of generating story plots from a case base of existing stories analyzed in terms of Propp functions. A CBR process is defined to generate plots from a user query specifying an initial setting for the story, using an ontology to measure the semantical distance between words and structures taking part in the texts.