This blog has been really helpful for me while I was writing my thesis about the uncanny valley. Now that I’ve finished my PhD, I thought it might be nice to include a potted synopsis on the valley and the work I’m doing. This is taken from an article I wrote for the OU’s student magazine in December 2012.
Studying The Uncanny Valley Stephanie.Lay@open.ac.uk
When I started researching the uncanny valley back in 2005, no-one knew what I was talking about. I’m now finishing my PhD and while it’s not yet a household term, it’s sufficiently well known that I’ve managed to collect over 1,000 responses to my latest piece of research. Today, I’d like to talk more broadly about my research and method of exploring the valley: my first task was to first consider the question of human-likeness. There are many ways to make something artificial seem more human: improved mechanics may make it move in a more fluid way, more detailed materials may make artificial skin look less like plastic and more natural. However, these changes tend to be large scale, and I wanted something I could manipulate minutely. The method I chose has been to morphing a range of artificial faces into matched human faces: this way, I can control the features that are blended from one into the other, the type of blends used and the number of steps taken to go from artificial to human. In this example, a doll has been morphed into an adult female, but I have also used robots, statues, animals and mannequins as my non-human starting points:
You might want to consider for yourself whether this sequence of faces does have an uncanny valley point: for me, the seventh face looks most unsettling… This effect is particularly striking when viewing the faces as an animation and you can see this here: http://www.youtube.com/watch?v=Fug_WX0J6ck
Once I’ve created these subtle variations in human-likeness, the next step is to present them to people and explore how they feel. My last study presented people with several images that varied in human-likeness, and asked them to undertake two tasks: firstly, they were asked how they would describe each person to allow someone else to find them in a crowded place, and then to say how they would feel if that entity was spending time in their home. The analysis of the descriptions came up with novel and interesting findings: the number of references to eyes as opposed to other features was strikingly across the different categories of human-likeness.
For me, the key issues still to be resolved do revolve around face perception, but within the wider uncanny research community, projects are looking a whether mismatches between realistic voices and almost-human faces may cause an uncanny sense of dissonance, the prevalence of an uncanny effect in other species, and how to manipulate human-likeness to make horror games as disturbing as possible! It’s certainly a lively area for research, and one where important questions are still to be answered.
To read more about my research:
Data collection site: http://www.uncanny-valley.co.uk/
Research journal: http://uncanny-valley.livejournal.com/
My gallery of uncanny faces: http://uncanny-valley.livejournal.com/pics/catalog/1514
Twitter feed: http://twitter.com/uncannyvalley