Brain Machine Interfaces are far from being used in games

Imagine playing your favourite game with the mere thought of your mind. That has been the dream of many sci-fi fans and has been seen in movies and books. In the last 10 years, research into brain machine interfaces (BMI) has provided some interesting studies.

For example, researchers in Netherlands at have managed to get two subjects to play a simple game of pong by reading their brain signals through fMRI.  Another group of researchers from the University of Pittsburgh and Carnegie Mellon managed to train a monkey to control a prothetic arm with his mind, and use it to grab and eat food.

So will we be mind fragging our opponents in Counterstrike and mentally micromanaging our zerg swarm in the near future?

No. There are many reasons these developments are a long way off, and one of the main reasons is the following: The methods used to read brain signals are either too inaccurate or too invasive.

Reading signals from the brain has been the staple for neuroscientists since the beginning of the field. However even with the many methods that we have to read the brain, each method has restrictions in terms of temporal and spatial resolution, invasiveness, and cost.

Electroencephalograms (EEGs) reads brain activity from electrodes placed on the scalp. EEGs are minimally invasive and have a good temporal resolution – meaning that there is very little delay between the measurment of the signal and the actual neural activity. The downside is that the signals have a very poor spatial resolution. The EEG signal reflects the activity of many millions of cells with very little specificity on where the signal is coming from. As such, using EEG signals would be very difficult, because understanding the meaning of the signals would be extremely hard.

Functional magnetic resonance imaging (fMRI) is another possible method. fMRI measures the changes in blood oxygen levels in the brain, which is associated with brain activity. As mentioned earlier, this method has been successfully used to play a simple game of pong. fMRI has a better spatial specificity than EEGs and is not invasive. However the downsides to this are that the fMRI signal has a low temporal resolution – the measured blood oxygenation response is occurs only 4-5 seconds after the corresponding brain activity. Four to five seconds may be trivial when you just need to move your paddle to hit a virtual ball, but if you want to snipe someone from across the map or engage your minions in battle you would need a better reaction time. Oh, I also should mention that the price tag for fMRI machines goes into the millions of dollars.

Finally there are single and multi-unit recordings. This method gets up close and personal to your neurons by inserting one or many electrodes directly into your brain. The advantages of these methods are they have both good temporal and spatial resolutions. The measurements can tell when and where your brain was active in the region that it is recording. In research, this method has been the most used. However, it is needless to say this method is extremely invasive. I am sure that you wouldn’t want to have your brain poked by a hundred little spikes for the sake of gaming.

So the non-invasive methods of brain recording (EEGs, fMRI) are not accurate enough, and the invasive methods (single and multi-unit recordings) are too dangerous. Actually, when I look at it, I think it is pretty ridiculous to even speculate the possibility of commercial brain machine interfaces.

Posted in Uncategorized | Tagged , , , , , | Leave a comment

Psychologists working in the video game industry

At the moment, I don’t think that there is a large demand for psychology majors in the gaming industry. This lack of demand may change, as developers seek to improve game immersion, gameplay mechanics, and understand social interaction between players.

I’ve done some digging to find two people with graduate degrees in psychology who make a living working at video game companies.

Mike Ambinder works as an experimental psychologist  at Valve Corporation and has worked on games such as Team Fortress 2, Left 4 Dead, and Portal 2. He holds a B.A. in Computer Science and Psychology from Yale University as well as a PhD in Psychology from University of Illinois at Urbana-Champaign. In his valve profile he describes his job description as “vague, but he thinks it probably has something to do with applying both psychological knowledge and methodologies to game design.” In March 2011, he gave a talk at the Game Developer’s Conference on biofeedback and game design.

Célia Hodent works as a playtest project manager at Ubisoft. She holds a PhD in cognitive psychology from the University of Paris Sorbonne where she studied cognitive development in children. She describes her role as “[helping] production teams to concretely integrate what we know about the brain from cognitive neurosciences into video games”.


Posted in Uncategorized | 2 Comments

Starcraft 2 Automated Player

Matt Fisher, a Stanford PhD student has revealed an incredibly impressive automated Starcraft 2 player. This is not just any other computer AI.

Screen capture of the AI's console output(Screen capture from Matt Fisher)

Some key features of the player’s AI:

  • The AI plays just like a human player: It looks at the screen, and then sends mouse and keyboard instructions.
  • The AI employs parallel processing: It maintains a series of threads which generate execution orders with a priority value. Keyboard and mouse control is then relinquished to the highest priority order at any given time.
  • The AI can perform an absurd amount of actions per minute (APM), ranging from 500 when idle to 1000-2000 during intense battles.

Finally, the AI can trash talk!

The AI maintains a set of personas it adopts when fighting humans. These are designed to intimidate the person with the AI’s ability to type very complicated and computationally intensive sentences in the middle of battles.

The AI successfully beats the normal SC2 computer opponents and some of Matt’s friends. However, he has yet to unleash it onto unsuspecting opponents on

I’m looking forward to seeing future developments!

Discussion thread on Hacker News

Discussion on Reddit

Posted in Uncategorized | Tagged , , , , | Leave a comment

Reward Anticipation in TF2 Crates?

Team Fortress 2, Mann. Co Supply Crates can be opened to reveal virtual items by buying a $2.49 key. This surprise item can consist of ordinary weapons to hats or even “exceedingly rare items”. Some of these rare items can fetch over $400 over the black market.

One of the interesting things about opening a crate is that there is a 5 second countdown. Below is an example:

Is there any significance to perposefully including a delay between you and your newest loot?

Robert Sapolsky, Professor of Neurology and Neurological Sciences at Stanford University may say yes. In a recent study, monkeys were trained to press a lever to receive a reward. They measured transient levels of dopamine in the monkey’s brain and found that dopamine spiked not during the reward, but during the anticipation of the reward. Furthermore, if the reward is only given 50% of the time, dopamine levels will increase further!

So maybe the game designers at Valve have realized this and put in that delay on purpose.

For more on this topic:

Reward anticipation – A powerful tool for game design – A Good overview by Lennart Nacke of Sapolsky’s talk as well as relevant examples from World of Warcraft.

Predictive Reward Signal of Dopamine Neurons – The original scholarly article about dopamine and reward anticipation.

The dopaminergic basis of human behaviors: A review of molecular imaging studies – A 2009 review paper on dopamine and human behaviour.


Posted in Uncategorized | Tagged , , , , | Leave a comment