Our goal is to understand the computing architecture of the human mind and brain. We bring the mind to its computational limits by putting it in its natural environment, where it needs to cope with survival threats. This requires rapid actions with utmost precision and continuous updating. To this end, our experimental platforms use serious games and virtual reality, as well as tightly controlled associative learning setups. Our theoretical models are informed by those used in artificial agents. Beyond understanding rational avoidance, we are also interested in psychiatric conditions that are characterised by avoidance actions in the absence of real threat, such as anxiety disorder or post-traumatic stress disorder. We seek to unpack the underlying mechanisms and develop novel treatments.