This is the presentation I gave in Solmukohta, Nordic Live Action RolePlaying convention (February 24th - 27th, 2000, Helsinki, Finland) as a part of "Roles, Avatars, Computers & Gaming session".
Computer programs called agents can be used to create emotionally intelligent NPCs for virtual game worlds and virtual realities. Agents are autonomous programs which can act own their own and do some tasks independently. By emotionally intelligent NPC I mean that NPC has personality, has emotions and it can show. NPC should also be able to learn to recognize places and other characters and form opinions about those. Personality of NPC should have an effect on what it does; goals and likes versus fears and dislikes defines action in any situation. NPCs do not need to be smart; they just need to seem to have an appropriate level of intelligence for that character. Thus NPCs are needed for rich and believable game worlds. Ian Wilson puts it:
Emotion primary serves social function in interactive entertainment. Emotional responses are used to make characters that we encounter believable and engaging. For example, if we were to walk into virtual bar and all the characters in the bar had distinct personalities the scene would be very impressive and believable social situation. (Wilson 1999.)
Life-like characters are needed in creating life-like game worlds.
Lets first take a look what there are today; Is there games or virtual worlds that are inhabited with characters that have a personality - that maintain an illusion of personality.
Computer role-playing or adventure games as I would rather call them, are currently series of puzzles that the player has to solve and then there are maybe some action sequences. Unfortunately game developers seem to be mostly focusing on implementing better AIs for combat and therefore social interaction is almost at the same level than in the best adventure games in 1980's. Interaction and dialog are strictly pre-scripted and they leave almost no alternatives to the player - and plots are almost always poor.
Chatterbots are programs that engage in conversations with natural language. They seem to have personality even they are not intelligent.
Eliza is a chatterbot developed in late 1960s at MIT. She communicates with simple techniques for interaction; she use text analysis, pattern matching and evasion technique to produce answers that seem like meaningful and intelligent. (Klein 1999 p. 28.)
Julia is chatterbot that connects to a MUD and besides engaging conversations she has simple memory on what has been said and where she has been. She is a descendant of Eliza. When she is not engaged in conversation she wanders across the MUD. Julia is also able to quote on player to another. She knows whereabouts of players, can flirt and talk about hockey etc. One feature is that Julia insists that she is human quite efficiently - she had convinced many players that she is human, at least for a while. (Turkle 1995.)
Erin the bartender is newer chatterbot. Erin serves drinks and chats with customers. Music is her favourite subject. She has emotional state which affects her responses. She also forms attitude towards her customers, and that too has impact on her responses. (Bates 1992.)
Oz project has been researching on believable agents that exhibits rich personalities and use natural language to communicate. Their focus is in interactive drama in virtual worlds inhabited by characters, believable agents - with whom audience can interact. Believability in this context means that character has:
And last but not least; Character grows and changes with time. This all means that everything character does it does it own personal style. As their focus is in interactive drama they also study story and elements of story, what makes story interesting and how interactive stories should be modelled. It seems that Oz-project is reinventing role-playing based on theatre. (Mateas 1997.)
Institute for Applied Intelligence has introduced a method for creating stories by using template stories and Oz-like character agents. Creating consistent, plausible and interesting story automatically is extremely difficult. They use affective reasoning mechanisms, a process model of emotions, to tag basic plot steps in stories with varied interpretations on behalf of each participating agent. Narratives with little constraints can be created by combining various interpretations of plot steps. This approach allows creating story-telling agents and actors that can be scripted ad hoc. So for example we could define different plot steps for bartender and customers of virtual tavern, and introduce wide variety of story-morphs in to the dialog. (Elliot et al 1998.)
Gaming is basically social. Solitaire games existed but they were rare exceptions before time of PC. Console games where almost always designed to be played by one or two persons. PC games introduced environments that forced games to be played as solitaire - one screen, one keyboard. (Costikyan 1998.)
Recently we have seen invasion of multiplayer PC games. Those are mostly multiplayer first person shooters.
Virtual worlds offers great environment for social gaming and role-playing. Technical attributes in VR have significant effect on social and cultural phenomenon what are happening in VR (Reid 1994). Fancy 3D graphics and special effects are not necessarily needed because VR [and traditional role-playing games] are more imaginative experience than sensory experience. Players are willing to maintain suspension of disbelief and let their imagination enrich the game world. Elizabeth Reid describes possibilities of text-based virtual realities MUDs:
Despite being textually based, MUDs are sites for social interaction and culture meaning. The virtual worlds created with MUD software are dramaturgically and socially rich, and MUD players have been able to device means of communicating social context cues through the textual medium. (Reid 1994.)
I believe that role-playing gaming basically doesn't differ much from gaming in VR and that gaming cultures can make more difference how a game is played than technology - tough technology has impact on gaming cultures in that media as said before. For example gaming in some game centred MUDs focus more or less on only killing mobiles, monster that game controls and gaining experience points. Some ways of playing tabletop role-playing games don't differ at all from that. Social codes of some MUD demand role-playing when interacting with other characters. You can even play LARP without role-playing the character by being yourself and focusing problem solving even when game mechanics of LARP supports the role-playing.
Mechanics of MUDs offer possibilities to be somebody else, not just play somebody else. Sherry Turkle writes in her book "Second Self" about identities in MUDs:
MUDs blur the boundaries between self and game, self and role, self and simulations. One player says, "You are what you pretend to beyou are what you play." But people don't just become who they play, they play who they are or who they want to be or who they don't want to be. Players sometimes talk about their real selves as composite of their character and sometimes talk about their screen personae as means for working on their real live lives. (Turkle 1995.)
MUDs are a form of VR and what Turkle said can be generalized to virtual realities, but not to traditional RPG or LARP. Therefore social interaction and role-playing can be used to make computer games more interesting and appealing. Note that not every good game needs those but as described earlier, game designers have successfully used this tactics in creating MUDs that has attracted players a long time. MUDs still attract new players despite the fact that they are basically same kind of game as first introduced in 1979.
I think that in the near future we could see virtual realities that are inhabited autonomous emotionally intelligent agents, NPCs, which performs routine tasks; such as working as a bartender, bouncer or city guard. Is should be able to chat with players, make friends or enemies, thus supporting role-playing and social interaction more than traditional virtual game worlds. Even mobiles could have personality and ambitions and so they could be in more important role in the world than currently - not just something to kill but be able to have social interaction and co-operation.
One example of game development that is heading in this direction is WorldForge project. They intent to implement, and have already done some of the work, AI engine that controls NPCs with ambitions and that support complex interaction Those NPCs has memory of things and places they have seen, thing that they have tried, knowledge of places and of other characters. Their short time goal is to make NPC understand simple language. (WorldForge.)
I think that next step would be to design games that use story-morphing techniques or believable agents to create freeform games with complex plots so that game starts to feel more like interactive drama or role-playing game than problem solving or killing contest; games that focus on social interaction and outcome of the game depends just on action of all characters - player characters and NPCs alike. Actions of players create the story and drama. So focus then moves to designing drama into the game using characters and their motivations and personalities; Exactly the same problem that writers of traditional role-playing games has been facing, but different perspective. For example Sokal's Canardo: "L`ile noyée" (Sokal 1992) could be easily be adopted as LARP, murder mystery weekend or computer game with interactive story. Plot would be perfect for a game adaptation because it has small number of characters and place where all happens is a small island - so it would be perfect set up for game with social interaction. Just putting characters in the virtual island would start the drama.
I will sketch the plot as interactive drama. Firstly the important characters of the cast:
Then we'll add couple of curious tourist and the owner of the only hotel of the island. We could use following plot steps to ensure drama in the story:
NPCs will try to follow plot steps and act according to their goals and personalities. In the game story could go many different paths depending interaction between NPCs and player character(s). It is likely that some plot steps will not happen in all variations. You might have noticed that plot description looks much like description of freeform RPG module.
Costikyan, G. 1994. I Have No Words and & I Must Design. http://www.crossover.com/~costik/nowords.html.
Costikyan, G. 1998. Why Online Games Suck (And How to Design Ones That Don't). http://www.crossover.com/~costik/onlinsux.html.
Elliot, Brzezinski, Sheth & Salvatoriello. 1998. Story-morhping in the Affective Reasoning paradigm: Generating stories semi-automatically for use with "emotionally intelligent" multimedia agents.
Excalibur. http://www.ai-center.com/projects/excalibur/.
Foner, L. 1997. Entertaining Agents: A Sosiological Case Study. MIT Media Lab.
Kelso, M. Weyhrauch, P. & Bates, J. 1992. Dramatic Presence. Mateas. 1997. An Oz-Centric Review of Interactive Drama and Believable Agents.
Reid, E. 1994. Cultural Formations in Text-Based Virtual Realities. http://people.we.mediaone.net/elizrs/cult-form.html.
Reeves & Nass. 1996. The Media Equation: How People Treat Computers, Television, and New Media Like Real People and Places, Cambridge University Press
Spector, W. 1999. Remodelling RPGs for the New Millenium. Game Developer Magazine, February, 1999. Saatavilla myös http://www.gamasutra.com/features/game_design/19990115/remodelling_01.html.
Turkle, S. 1984. Second Self. Simon and Schustler. New York.
Turkle, S. 1995. Life on the Screen: Identity in the Age of the Internet. Simon and Schuster, New York.
Twilight Minds. 1999. Brainiac Behavior Engine. http://www.twilightminds.com/bbe.html.
Wilson, I. 1999. Artificial Emotions. http://www.gamasutra.com/features/19990507/artificial_emotion_01.htm.
WorldForge. http://www.worldforge.org/.