According to Wikipedia, cognitive biases "are tendencies to think in certain ways that can lead to systematic deviations from a standard of rationality or good judgment, and are often studied in psychology and behavioral economics." Far more than simply an exercise in academia, cognitive biases have all sorts of practical impacts on our lives, whether or not we admit it. A very broad umbrella, cognitive bias comes in many forms, as evidenced by the fact that Wikipedia lists over 170 of them. Some of these biases are more prevalent in certain areas of life than in others. Below is an infographic from Business Insider, of all places, which is an elementary summary of what it refers to as "20 cognitive biases that screw up your decisions." But how do these cognitive biases relate to real life?

A cognitive model of human learning provides information about skills a learner must acquire to perform accurately in a task domain. Cognitive models of learning are not only of scientific interest, but are also valuable in adaptive online tutoring systems. A more accurate model yields more effective tutoring through better instructional decisions. Prior methods of automated cognitive model discovery have typically focused on well-structured domains, relied on student performance data or involved substantial human knowledge engineering. In this paper, we propose Cognitive Representation Learner (CogRL), a novel framework to learn accurate cognitive models in ill-structured domains with no data and little to no human knowledge engineering. Our contribution is two-fold: firstly, we show that representations learnt using CogRL can be used for accurate automatic cognitive model discovery without using any student performance data in several ill-structured domains: Rumble Blocks, Chinese Character, and Article Selection. This is especially effective and useful in domains where an accurate human-authored cognitive model is unavailable or authoring a cognitive model is difficult. Secondly, for domains where a cognitive model is available, we show that representations learned through CogRL can be used to get accurate estimates of skill difficulty and learning rate parameters without using any student performance data. These estimates are shown to highly correlate with estimates using student performance data on an Article Selection dataset.

Over the last few months, I've been catching up more systematically on what's been happening in machine learning and AI research in the last 5 years or so and noticed that a lot of people are starting to talk about the neural net developing a'mental' representation of the problem at hand. As someone who's preoccupied with mental representations a lot, this struck me as odd because what was being described for the machine learning algorithms did not seem to match what else we know about mental representations. So I've been formulating this post when I was pointed to this interview with Judea Pearl. "That sounds like sacrilege, to say that all the impressive achievements of deep learning amount to just fitting a curve to data. From the point of view of the mathematical hierarchy, no matter how skillfully you manipulate the data and what you read into the data when you manipulate it, it's still a curve-fitting exercise, albeit complex and nontrivial."

There are 4 qualities of the universe that limit our own intelligence and the intelligence of every other person, collective, organism, machine, alien, or imaginable god. All 200ish of our known biases are attempts to work around these conundrums! The 1st conundrum is that there's too much information in the universe for any individual within the universe to process it all. We have our 5 senses (or up to a dozen depending on how you divide them up), and we're located at points within vast planes of space and time. So there's a lot of information out there (outside your house, across the street, on the other side of the world, throughout the galaxy, and back in time) that we have missed and will continue to miss.

This paper draws a parallel between similarity-based categorisation models developed in cognitive psychology and the nearest neighbour classifier (1-NN) in machine learning. Conceived as a result of the historical rivalry between prototype theories (abstraction) and exemplar theories (memorisation), recent models of human categorisation seek a compromise in-between. Regarding the stimuli (entities to be categorised) as points in a metric space, machine learning offers a large collection of methods to select a small, representative and discriminative point set. These methods are known under various names: instance selection, data editing, prototype selection, prototype generation or prototype replacement. The nearest neighbour classifier is used with the selected reference set. Such a set can be interpreted as a data-driven categorisation model. We juxtapose the models from the two fields to enable cross-referencing. We believe that both machine learning and cognitive psychology can draw inspiration from the comparison and enrich their repertoire of similarity-based models.

I tried to show that in my post. One problem is the current inputs lose a lot of the information that human mental representations draw on. But NNs only work on vectors represented as numbers.

Two years ago, Google made an internal video that didn't stay internal for long. Recently acquired by The Verge, it tells the speculative story of how the technology giant might develop a universal model of human behavior by collecting as much data from people as possible. The video, titled "The Selfish Ledger," is a thought experiment that shows how a major institution like Google could make use of the complex data profile built up by each person as they buy, browse, and communicate online. Then in true form to tech monoliths' disregard for data privacy, the video suggests the following: What if the ledger could be given a volition or purpose, rather than simply acting as a historical reference? What if we focused on creating a richer ledger by introducing more sources of information? What if we thought of ourselves not as the owners of this information, but as custodians?

We are on the verge of another revolution in health care: deeply personalized medicine. It's the next computerized step in tailoring medical treatments and medical drugs to your specific body, your very unique anatomy, the specific ways your body works and doesn't, and your path to live your life and keep healthy. But we may soon run into problems of ethics and personal privacy that could make the recent furor over Facebook and data mining look small by comparison. Personalized health and wellness comes from the intersection of improved body-worn sensors, data science, computational physiology, individually customized health assistance and -- if necessary -- highly targeted medical treatment, all coming together at once. This dramatic health revolution is enabled by the vastly reduced cost of reading and analyzing our genomes, and of huge, cheap quantities of computer power that allow us to make more precise predictions about our future health from our genetic setup.

A new algorithm in artificial intelligence enables a 3D model of a person to be created in just a few seconds after videoing their features. Artificial intelligence is used during video games and virtual reality to create 3D objects of people and objects. But typically it requires special equipment when filming in order to transfer the video of someone into a 3D figure. New video software is able to take the footage and transfer it into the model in seconds from just one angle. A minute-and-a-half long video shows how the algorithm is able to transform the images of men and women into a 3D character after they turn around themselves, Science Magazine reported.

Transporting yourself into a video game, body and all, just got easier. Artificial intelligence has been used to create 3D models of people's bodies for virtual reality avatars, surveillance, visualizing fashion, or movies. But it typically requires special camera equipment to detect depth or to view someone from multiple angles. A new algorithm creates 3D models using standard video footage from one angle. The system has three stages.