The Uncanny Narrative Valley.

Procedural narrative has its own Uncanny Valley and we should be aware of it in order to avoid or minimize its effects.

I love to use other fields terminology when I speak about narrative design.

Narrative on videogames is a very undeveloped sector. When we talk about videogames there is always the adage of “gameplay is king, narrative shouldn’t get in the way” so there is a lot written about gameplay design but there is little documentation about narrative, leading us to use narrative methods borrowed from other mediums like cinema, television or literature where we have studies, analysis and proven tests about the things that works and the things that don’t.


The absence of a clear golden path is usually confused with an invitation to anything goes but the reality is that we have a lot of techniques already proven on different areas that can provide useful perspective to the narrative tasks.


I like to compare the interactions between narrative elements with collisions on a physics system, how we should measure interactions intensity in order to provide different responses, the same way that we don’t make a box fly the same distance when slightly pushed with the hand than when it’s shot with a cannon, or how we should extend the narrative consequences to future interactions between NPCs without the player inference the same way pool balls interact between them way after the white ball gave the first hit.


I find useful to speak about Level of Detail on narrative because the same way we provide the player a more or less detailed model based on the distance to it, to optimize the performance, we can provide more complex characters with more detailed stories based on the “emotional distance” between the player and those characters.


And the same way that graphics and animation have the now famous “Uncanny Valley”, I tend to use the name “Uncanny Narrative Valley” given to a situation that happens when we make overuse of procedural methods on narrative.


Just to put ourselves on context, the Uncanny Valley refers to a situation when a CGI character looks almost human but we can’t stop focussing on how much it fails to look totally human. It is a weird situation because our brains don’t miss a beat when giving human qualities to characters that are obviously not human. We usually have no issues on considering that Lightning McQueen, C3PO or Mickey Mouse have human feelings and thoughts, but we cringe when we see the ultra realistic models on the Final Fantasy movie or when CGI is trying to fake us into thinking we are seeing a young Princess Leia.


When we create the narrative on a game, there is a limit on the amount of content we can possibly generate by ourselves. We can’t handmade create as many lines for a character to say as the possibly number of times a player will be willing to try a new conversation and we can’t create as much situations as players will be possibly eager to go through, so we can start using procedural generation to help us.


We can create a certain amount of content and create combination rules so the content can mix and match. That way, if we have five ways a story can start, five ways you can find the first clue, five possible turnpoints and five possible outcomes and we define that a storyline will use one start, one way to find the first clue, one turnpoints and one resolution, we can see, on a raw approach, that we can create 5x5x5x5 = 625 different storylines with just 40 handmade pieces. A different question is if we can guarantee that those 625 results are equally valid, but for the sake of the argument let’s pretend that they are.


Now we have all this content that we can use, but the question is: Should we use it? Cue in the Uncanny Narrative Valley.


The brain is specialized on pattern matching. This works on two directions… we feel comfort when we sense that everything is as it should be and we unnerve when our brain thinks something’s wrong, even if we don’t know what. On the case of the CGI Princess Leia everything seems perfect but there is something… how she moves, how the light reflect… we don’t know what, but we notice, we don’t like it and we forget about all the hard work behind that incredible model and animation. From that moment on we can only see that she is not a human being. This also happen with procedural narrative.


Perhaps we don’t notice on the first occurrence but, after a small amount of repetitions, even with totally different storylines, without repeated handmade content, somehow we start noticing the seams and from that moment on we fall on the uncanny narrative valley.


Perhaps a totally handmade game doesn't offer such a great number of storylines with such amount of complexity, but it feel “human”... you start forgiving the faults because you empathize with the creator, you feel the work.


On the uncanny narrative valley suddenly all the hard work and handmade content stop being meaningful. We can only focus on how the storyline is generated, where a piece ends and the next starts, and the more we notice the more bitter we feel about how a computer was trying to trick us.


There are several approaches to try to minimize this already being tried. One is, obviously, more handmade content with different “stitches”. Add storylines with a different number of turnpoints, without the first clue piece because it is directly given on the start, etc. The more ways we can recombine, the longer it should take us to start noticing. Another one could be providing an explanation about the repetition pattern. Perhaps your character is on a temporal loop, or he is facing a serial killer that always use same modus operandi. A third one can be no stitching at all. Make everything happen as individual events and let the player connect the dots. A different approach is using playtesting to estimate how long we can make use of one pattern before it starts being noticeable and then change to a different one. This last one can be tricky as different people have different sensibilities and start noticing at different paces...


This is not a call to stop using procedural narrative, on the contrary, I encourage everyone on using it, experiment with it and try to find solutions to this issue that I think is one of the most usual shortcomings on games that make use to certain extent of procedural techniques to generate narrative content.

Latest Jobs

IO Interactive

Hybrid (Malmö, Sweden)
Gameplay Director (Project Fantasy)

Arizona State University

Los Angeles, CA, USA
Assistant Professor of XR Technologies

IO Interactive

Hybrid (Copenhagen, Denmark)
Animation Tech Programmer

Purdue University

West Lafayette, IN, USA
Assistant Professor in Game Design and Development
More Jobs   


Explore the
Advertise with
Follow us

Game Developer Job Board

Game Developer


Explore the

Game Developer Job Board

Browse open positions across the game industry or recruit new talent for your studio

Advertise with

Game Developer

Engage game professionals and drive sales using an array of Game Developer media solutions to meet your objectives.

Learn More
Follow us


Follow us @gamedevdotcom to stay up-to-date with the latest news & insider information about events & more