From Psychohistory to Babelfish – From predicting the future to mind-reading
My research of the last twenty years was inspired by Isaac Asimov’s fictional science of psychohistory. In his Foundation Series science fiction stories, Asimov describes how mathematician Hari Seldon was able to predict the future a thousand years ahead through the discipline of psychohistory that Seldon invented. Its premise was that while it is impossible to predict the behavior of an individual, the aggregated behavior of millions and billions of people extending over the galaxy can be predicted accurately by studying their communication patterns. However already in his stories Asimov posited that it is impossible to predict “black swan events”, i.e. unexpected random events. In Asimov’s story, the appearance of the “Mule”, a mutant conqueror who could read and manipulate the mind of others, a few hundred years after Seldon made his predictions, threw Seldon’s predictions totally off.
This is where Babelfish comes in, a tool envisioned by Douglas Adams in his science fiction classic “The Hitchhiker’s Guide to the Galaxy”. If you slide a Babelfish into somebody’s ear, they will be able to understand any language of anybody from anywhere in the Universe. Babelfish does this by directly reading the brainwave matrix of the counterpart, and beaming it into the mind of its wearer.
In my psychohistory-inspired research over the last twenty years, I have used online social media and email to predict external events, for instance predicting stock markets from Twitter, or the outcome of Academy Awards from online forums, or performance of organizations from their email archives. However, just like in Asimov’s psychohistory, while we were able to perfectly predict the past, i.e. fitting our regression and AI machine learning models to past data with high accuracy, and then claiming that the same prediction model would also be true in the future, any unexpected black swan event threw off our prediction model. For instance, the election of Donald Trump, or Covid-19, nobody saw them coming, and it invalidated any stock trend prediction model running at that time. Another black swan event are the changes Elon Musk imposed on Twitter, now renamed X, after he acquired it. This means that any elections/stock trends/epidemiological/etc. model based on Twitter will have to be rebuilt from the ground up.
Motivated by the deficiencies in psychohistory, I started building in baby steps tools resembling “little Babelfishes” for the last six years, trying to read the minds of individuals from their words, body language, and other “honest signals”. Just like Asimov’s “Mule” through his mind reading skills was able to conquer the universe, throwing off safeguards put into place by Hari Seldon through the mathematics of psychohistory, our research tries to read what people are REALLY thinking – in spite of what they say. We use AI-driven natural language processing (NLP), face emotion recognition, and other body signal recognition to build a series of tools to read the minds of people, animals, and plants.
Reading the mind - Communicating with people, animals, and plants through AI
If you stick a Babelfish in your ear, you can immediately understand anything that has been said in any form of language by any alien anywhere in the galaxy. The Babelfish decodes the brainwave matrix of your counterpart and transmits their intent directly to your mind. Similar to Babelfish, our goal is to build tools leveraging the latest advances of today’s AI to better communicate with animals, plants, and other people.
How can we better communicate with other people?
Just like Babelfish, we try to read what the other person is REALLY trying to say. Spoken language is just a small part of it, albeit the one which is the easiest for others to (mis)interpret. But humans also speak for instance with facial expression, body posture, gaze direction, and pheromones.
Our approach consists of reading the emotions of other people. It analyses their personality characteristics from their word usage, social networking dynamics, facial expressions, and body language. Towards that goal we have developed our own personality model, made up of different personas (personality archetypes). We developed the groupflow personas (bee/ant/leech) and alternative reality personas (fatherlander/nerd/spiritualist/treehugger). We also compute the FFI characteristics (openness, conscientiousness, extroversion, agreeability, neuroticism), Schwartz values (tradition, achievement, power, benevolence), Haidt moral foundations (care, fairness, loyalty, authority, sanctity), and DOSPERT risk attitudes (financial, health, recreational, ethical, social risk taking) from the words people use, but also from their facial expression in response to provocative movies, or simply by observing how they move their body, or the tone and pitch of their voice.
Our tools provide both a lens and a mirror. Similar to the magic mirror of the evil stepmother of Snow White, we provide a virtual mirror to a person, that tells them their beauty in comparison to others. It tells them what others think about them and how much they trust them, so they can improve themselves. It also provides a lens to look at others to gain insights about how they see me.
While we have focused for the last twenty years mostly on the dynamics of online interaction using text and online social media, our team recently has developed a series of new video and audio analysis tools to offer a magic mirror for face-to-face interaction:
The objective is to build and test a real-time magic mirror that analyses teams collaborating face-to-face, and gives team members real-time feedback on the quality of their interaction, and how they can improve their teamwork to get into groupflow. It combines the tools and methods (1) to (4) described above. The goal is - as our and the research of many others has shown - to become a better team player by being more conscientious, caring, collaborative, and creative!
How can we better communicate with animals?
What if we could build a Babelfish to talk not just to humans, but also to animals? In fact, at least in humble beginnings, we can. We can apply similar AI algorithms we developed to read human emotions to reading emotions of animals. In past research we have measured the emotions of dogs, cats, horses, and cows based on their facial expression, body posture and voice.
In further work we want to extend this to reading the intents of animals (defining a list of actions the animal wants to do), analyzing for example interaction between a dog and a human, or two dogs solving a task. This will give a similar magic mirror to dog owners interacting with their dog, helping them to communicating their intent to the animal by getting immediate feedback.
How can we start to communicate with plants?
Until recently, the vegetative nature of the plant as laid out over 2000 years ago in Aristotle’s treatise “On the soul” was recognized as accepted wisdom. But recently, it has been shown that far from just reproducing and growing, plants show awareness of their environment, communicating with numerous senses. Tomato plants produce sounds when stressed, mustard plants respond to the munching sounds of caterpillars by producing defensive chemicals, mimosas learn not to respond to harmless shaking, to name just a few examples. In our own research we measure the potential differences of plants such as basil or mimosas between their leaves and their roots. This is similar to the action potentials occurring in the brains of animals, which are responsible for the signaling of information in the body.
In our work we are using the plant spiker box to measure “emotions of plants”, i.e., potential differences in response to interaction with humans. In our work we have trained machine learning models to predict emotions of humans interacting with basil, mimosas, and garden plants. In new work we are building machine learning models which will, based on the plant spikerbox, recognize human movement, emotion, and different individuals.
We still have a long way to go from psychohistory to Babelfish. But if everybody eventually will wear a Babelfish, psychohistory might even work!
Comments
Post a Comment