Representational support for narrative-based game creation

many different elements which are woven together when the game is played to allow the story to unfold in one of potentially a number of ways as the player explores the game world. A branching narrative can quickly get unwieldy because each choice point brings additional possible paths (as illustrated in Figure 2.5, Section 2.3.3). Whilst working in a game creation toolset, the disparate elements of conversations, character appearances and behaviours, the design of locations and the placement of objects and characters are all represented separately.

As discussed in 2.2.2, external representations are crucial for writing (Sharples 1996), but no representational support is currently available for creating and managing interactive, multimodal narratives within a computer game.

When creating a narrative-based game there is no equivalent of reading back over what has just been written. Instead it is necessary to exit the toolset, load up the game and play through it: a laborious process (more akin to the cycle of computer program editing and running) which can only be done intermittently. This can interrupt the flow of writing as the ‘reading’ of a game is necessarily detached from composing and revising.

With a commercial game, a team of writers will often be involved and they will use techniques similar to those used by filmmakers, making use of representations such as storyboards and dialogue scripting to support the storytelling activity as required (Handler Miller 2004; Bateman 2007). However, these are experts who have already developed the skills in question, and young people with no previous experience or skills in the area are likely to require additional support.

Keeping track of information about the narrative elements within a game is a complex task, and difficulties in managing this could make it harder for young people to develop the associated writing skills which will be so potentially useful to them. Additional representational support for storytelling may allow designers to get a better sense of how their game narrative is developing while they are in the act of creating it, helping them to create a better underlying story, convey it more effectively through narrative discourse within the game, and potentially,

gain a deeper understanding and command of the multimodal and interactive writing skills they are using.

As discussed in 2.2.2, the traditional writing process can be supported through external representations used by the author as part of the planning, composition, review and revision stages. The choice of representations in software interfaces is always important, particularly when building interfaces for educational software. These interfaces have to support the user’s learning as well as ensuring that interaction is straightforward and intuitive.

There are some representational support tools available designed to help skilled adult designers of interactive stories and virtual reality environments. Scenejo provides a graphical authoring system for managing interactive dialogues which make use of a chat bot, and includes a high-level plot graph for defining the overall structure of the dialogue (Weiss, Müller, Spierling et al.

2005; Spierling, Weiß and Müller 2006). VRBridge aims to make the authoring of virtual reality environments accessible to end-users by providing a series of useful representations to support the design task (Winterbottom, Blake and Gain 2006; Winterbottom and Blake 2008). Whilst these are not aimed at young people, and do not have learning goals around writing, their representational choices are nonetheless interesting and are discussed further in relation to specific design decisions in 3.4.5 and 3.4.6.

In the remainder of this sub-section, higher-level representational theory is examined. It is generally accepted that a single representation is not sufficient to represent all aspects of any complex entity (Schwarz and Dreyfus 1993), and a 3D computer game is certainly such an entity. It is clear then, that any given representation of a computer game will highlight certain aspects of the game over others. The match-mismatch hypothesis (Gilmore and Green 1984) states that where a representation highlights a certain type of information, tasks using that type of information will be easier to perform than those requiring other types of information.

Where the required information is implicit in a representation and needs to be inferred, the task will be harder than if the information was presented explicitly. Self-report by a group of children who used the NWN2 toolset over a week long workshop indicated that 3D area design was the activity on which most time was spent (Robertson and Good 2004). Whilst 3D area design is a very important part of creating a narrative-based game, it may not be desirable to emphasise this aspect of the task over all other tasks relating to narrative creation. Certain elements of narrative can be conveyed visually, but many others rely on events and interactions.

If only the visual aspects of the area design are represented, it may be hard for young designers to keep track of the interactive elements which can reveal embedded narratives and allow players to take part in enacted narratives within the game.

At present, when trying to write an interactive story, users of the NWN2 toolset have to keep an internal representation of the (potentially complex) branching plot in their minds. Trying to hold a mental representation in working memory over a period of time can lead to cognitive

overload, something which is to be avoided in learning environments (Mayer and Moreno 2003). Creating an interactive narrative can place a huge load on the user and may

understandably lead them to focus on other aspects of game creation which are better supported.

Providing a representation of narrative structure could support cognitive off-loading and allow more complex narrative elements to be built up. This is likely to be needed in addition to rather than instead of the existing representation.

Making a task easy to perform by providing explicit representation of the required information and supporting cognitive offloading when working on complex tasks is beneficial. However, making a task as easy as possible is not always the aim in educational software. There are a number of tasks which the current NWN2 toolset interface makes very quick and easy, including the creation of an area, and adding new characters, objects and scenery to the area.

Whilst it may seem beneficial for composition related tasks such as these to be easy to carry out, easier is not always better when it comes to educational tools. When the aim is to help students to learn, having an interface which make certain tasks too easy can work against this aim (Brna, Cox and Good 2001). For example, in the NWN toolsets the user can ‘create’ a character simply by dragging a character model title from a list into the 3D area. A new character now exists in the game world, but it is a generic character with default options chosen without any input from the user. Creating a character is an important skill in creative writing, but the toolset makes this an insignificant process. The user has no opportunity to practice developing a well-rounded character, or to reflect on the character’s motivations or back story. The task is only ‘made easy’

by hiding the complexity with which the user needs to engage if they are to develop skills in creating characters.

Norman distinguishes between experiential and reflective cognition (Norman 1993), and expresses concern that using multimedia learning environments can encourage experiencing when one should be reflecting. Experiential or ‘reactive’ cognition does not require deep thought and is event driven with automatic reactions following from input. Reflective cognition requires much deeper thought and tends to be slower and more laborious. It also requires the ability to store temporary results and use those results in further thought processes. For this reason external representations can facilitate reflective cognition by allowing more complex chains of reasoning to be built up, providing further evidence that representational support should be given for complex composition tasks such as structuring an interactive multimodal narrative.

Choice of representation can completely alter the mode of cognition used in a task. Svendsen (1991) concluded that whilst direct manipulation interfaces, in which tasks are often achieved by dragging and dropping graphically represented elements, can be very user-friendly, they can actually hinder problem solving if they are supportive of thoughtless action. A number of other studies have also indicated that direct manipulation interfaces are not always desirable in educational applications (Trudel and Payne 1995; Golightly 1996; Holst 1996; Rappin, Guzdial, Realff et al. 1997). This effect has been explored in relation to assistive interfaces which

externalise task relevant information in an attempt to improve usability (e.g. greying out inappropriate menu options) (van Nimwegen, Burgos, van Oostendorp et al. 2006). This

research indicated that such representational assistance can be counterproductive where the goal is learning, and led the authors to suggest that designers of educational systems consider

intentionally making interactions ‘more difficult’ or ‘less assisted’. So, whilst it is important to represent the relevant information over the irrelevant, it is also crucial that tasks are not made too easy through representational support; the learner needs to engage in considerable reflective cognition in order to develop the desired skills.

Although most work in this area has focused on direct manipulation, it is really only a specific case in a more general model of how interfaces can support learning. In each of these studies it is not direct manipulation itself which is the problem, but the tendency for direct manipulation interfaces to encourage unreflective approaches to problem solving, such as trial and error. In each example the direct manipulation elements of the interfaces allow users to rely on interface feedback to achieve tasks, placing them in a reactive mode indicative of experiential cognition rather than the deep thought mode required for reflective cognition. However, this is not a necessary characteristic of direct manipulation. In fact, it is the level of granularity at which the direct manipulation takes place which is key. Svendsen’s study uses a command line tool as the non-direct manipulation condition, but an onscreen keyboard tool which allows characters to be dragged in one-by-one could allow typing in a command line through direct manipulation.

However, because the level of granularity would be so low, the user would be unlikely to adopt the unreflective trial and error approach which hindered the problem solving (although the interface would clearly be problematic in a number of other ways).

The important distinction is between selection, which is quick and easy and can be done without deep thought, and composition, which requires considerable thought and effort. Direct

manipulation interfaces often limit users to selecting between one of a given number of choices, as in Svendsen’s study, but this is not necessarily the case and direct manipulation interfaces can support composition tasks if the granularity at which the manipulation takes place is small enough.

The key consideration for educational tools is that extensive reflective thought should be applied by the learners to the right activities. A game creation interface which made the task of saving the game into a reflective activity which required deep thought would be unsuccessful (unless the aim was to teach about the underlying process of writing a game file to disk).

Similarly a complex scripting language is an undesirable way of creating in-game events (unless the aim is to teach coding skills). As our aim is to develop writing skills, the interface should encourage deep reflective thought when it involves the practice and development of those skills, which are naturally composition-based tasks. Other aspects of game creation should be carried out easily through selection to avoid distraction and unnecessary cognitive effort.