Smith and Colantonio wanted to create a system of "Guiding and Attracting," encouraging players instead of forcing them to see certain places or features. Accordingly, they created a "Pull-based Narrative" with the story, empowering players while suggesting direction.
In order to pull off this vision, they sought to create "general purpose systems," environments in which entities – be they enemies or grenades – can influence each other "through an input/output system." As they describe it, they were making "game mechanics that 'listen' to each other." For example, they chose to allow "Summon Rats" to summon AI-controlled creatures into the world instead of limiting "Summon Rats" as an ability that players must use on enemies.
Fundamentally, general purpose systems allow for player creativity, which was an absolutely crucial element that made Dishonored so fun.
Smith did discuss some downsides to this approach. Some non-dramatic moments can occur. Additionally, "players are responsible for creating fun," something many games never have the audacity to empower players in that way. Of course this approach also takes a lot of work, fixing a myriad of unexpected bugs and developing all sorts of interesting features and relationships within the system. One particularly interesting early exploit included guards that would not notice when another guard was fried by the Wall of Light and charge through anyway, creating silliness where none should exist.
The Play-Path Matrix, a term they used to describe the ability for players to make many decisions within a given space, was also crucial for Dishonored's design. They wanted to make the game less predictable, hence the decision to randomize much of the game, including the location of targets. This fluid design approach is beautifully highlighted in Lady Boyle's Last Party, my personal favorite mission in Dishonored. As Smith explains, "this is an interactive form of drama."
Q: How did you decide on a certain number of abilities in the game?
A: Arkane prototyped about fifteen abilities and kept ten. The most fun systems win after rigorous testing.
Q: Is there a way you visualized emergent behavior to make sure it was all good?
A: Arkane encouraged emergent gameplay but never mapped out everything. Although they do have plenty of experience with this type of thing and did have discussions about what could happen. Colantonio said they planned about 10% of it.
Q: Considering how much trust you put in players, why would responsibility in players' hands be a downside of your approach?
A: Colantonio: You still need to add some guidance, a great level of tutorial. Distill these systems one at a time, so you don't overwhelm the player.
Q: How do you deal with the complexity of testing?
A: Smith: We never got bored of System Shock, so you put a lot of QA on it. Scripted sequences were probably more troublesome for the team than emergent stuff. You spend a lot of time playing the speculation game.
Q: How much of the narrative was planned in advance vs. waiting to see people play?
A: Colantonio: While the big picture was planned, the team changed the big picture when players found the game too complex. Smaller things, like Emily's drawings, come on the fly. Roughly 20% of the time.
Q: What steps did you take to communicate the opportunity of choice?
A: Smith: The team started super subtle, and constantly tried to be subtle, but had to be more clear as it progressed. In terms of choice, we deliberately tried to obscure it.
Q: How did you do a consistent game with two different locations?
A: Colantonio: It's really one company, so much of the organization and culture is the game. Arkane also used concurrent video conferences in both offices that were on all the time.