Making Robots Cry

Making Robots Cry

March 7, 2011

I read a journal article recently that said if you want your robot to make good decisions, you have to make it cry. Just the headline alone was intriguing to me. Robots with emotions? Tell me more!

It turns out, the researchers in the MIT Artificial Intelligence Lab wanted to improve AI agents’ ability to make decisions. (Because what good is artificial intelligence if it can’t decide whether or not to take over the world?) Drawing on emerging research from the neurosciences, these researchers were leveraging breakthrough information about the relationship between emotion, memory and decision-making in humans and attempting to apply it to their robot creations.  Because helping clients understand their customers’ connections between emotion, memory and decision-making is kind of my thing, I was doubly intrigued.

Over the past 20 years, social scientists have been learning more and more that decisions are not purely rational exercises

Decisions are precipitated and driven by non-conscious processes that are rooted in the relationship between memory and emotion. It goes a little something like this: When we have an intensely negative experience, like a child staring into the jaws of a snarling dog, our brains record and catalog the memory for future reference. In this case, the child might always be a little wary of the canine species because her brain will use that memory to provide protection for her in the future. The next time this child sees an approaching dog, her brain will retrieve the memory, she’ll feel the original emotion of fear, and she will protect herself from the animal.

The way the brain catalogs and retrieves memories is through emotion.

Much of what we’ve learned about the relationship between memory, emotion and decision-making came about with the development of functional MRI scans of the brain, as well as by studying people who have experienced brain injuries. By examining the way they function with the absence of certain areas of brain tissue, we can learn about how the rest of the brain works. Joseph LeDoux describes one example of this kind of research relating to memory and decision-making in his book, The Emotional Brain. He relates the story of a patient whose brain injury eliminated her ability to formulate short-term memories. Each time her physician entered her room, he would have to re-introduce himself, as she had no recollection of who he was. That part of her brain simply wasn’t working. Either out of genius or pure meanness, this physician decided to test the depth and breadth of her amnesia.

The next time he entered her room, he introduced himself and shook her hand like always. Upon shaking his hand, she immediately withdrew hers because he had placed a thumbtack inside his own. When they shook hands, she was pricked by the tack and experienced pain. The interesting part is, the next time he entered her room, she did not remember him, as usual, and he had to re-introduce himself. However, she would NOT shake his hand. While she had no memory of meeting him, she had learned to feel the emotion of fear when presented with the prospect of touching him. She could not articulate why, but we now know her brain non-consciously archived the negative experience from the thumbtack and retrieved it in order to protect her from pain. LeDoux says, “[The physician] had come to signify danger. He was no longer just a man, no longer just a doctor, but had become a stimulus with a specific emotional meaning.” (p.182)

To understand the real motivations for decision-making, we have to look into memory.

This is what fascinates me about studying decision-making and the reasons why people really do the things they do. We can try to explain what drives our behavior and our decisions, but honestly, sometimes we just don’t know because the influencers live deep down inside our emotional brains.

And this is the very reason why the researchers at the MIT Artificial Intelligence Lab wanted to help their robots feel emotion. By helping them to experience synthetic emotions, they could then help the robots store those experiences as memories to be retrieved later for decision-making. The approach used by these scientists creates a great model for marketers, as well. If, as marketers, we truly want to understand why our consumers buy what they buy and do what they do, we need to look at the mental models they have developed for our products and our brands. Negative (and positive) emotional experiences with products and brands work in the brain the same way as dog bites and thumb tacks, although hopefully less scary and painful. By understanding consumers’ memories and emotions from the past, we can gain a better perspective on what they will do in the future.

Subscribe to our Blog