Reinforcement Learning from Human Feedback (RLHF) is a methodology that aligns agent behavior with human preferences by integrating human feedback into the agent's training process. We introduce a possible framework that employs passive Brain-Computer Interfaces (BCI) to guide agent training from implicit neural signals. We present and release a novel dataset of functional near-infrared spectroscopy (fNIRS) recordings collected from 25 human participants across three domains: a Pick-and-Place Robot, Lunar Lander, and Flappy Bird. We train classifiers to predict levels of agent performance (optimal, sub-optimal, or worst-case) from windows of preprocessed fNIRS feature vectors, achieving an average F1 score of 67% for binary classification and 46% for multi-class models averaged across conditions and domains. We also train regressors to predict the degree of deviation between an agent's chosen action and a set of near-optimal policies, providing a continuous measure of performance. We evaluate cross-subject generalization and demonstrate that fine-tuning pre-trained models with a small sample of subject-specific data increases average F1 scores by 17% and 41% for binary and multi-class models, respectively. Our work demonstrates that mapping implicit fNIRS signals to agent performance is feasible and can be improved, laying the foundation for future brain-driven RLHF systems.

COVID-19 exposed the need to identify newer tools to understand perception of information, behavioral conformance to instructions and model the effects of individual motivation and decisions on the success of measures being put in place. We approach this challenge through the lens of serious games. Serious games are designed to instruct and inform within the confines of their magic circle. We built a multiplayer serious game, Point of Contact (PoC), to investigate effects of a serious game on perception and behavior. We conducted a study with 23 participants to gauge perceptions of COVID-19 preventive measures and quantify the change after playing PoC. The results show a significant positive change to participants' perceptions towards COVID-19 preventive measures, shifting perceptions towards following guidelines more strictly due to a greater awareness of how the virus spreads. We discuss these implications and the value of a serious game like PoC towards pandemic risk modelling at a microcosm level.

Natural Language Processing (NLP) systems commonly leverage bag-of-words co-occurrence techniques to capture semantic and syntactic word relationships. The resulting word-level distributed representations often ignore morphological information, though character-level embeddings have proven valuable to NLP tasks. We propose a new neural language model incorporating both word order and character order in its embedding. The model produces several vector spaces with meaningful substructure, as evidenced by its performance of 85.8% on a recent word-analogy task, exceeding best published syntactic word-analogy scores by a 58% error margin. Furthermore, the model includes several parallel training methods, most notably allowing a skip-gram network with 160 billion parameters to be trained overnight on 3 multi-core CPUs, 14x larger than the previous largest neural network.

Road sign recognition is a key component in autonomous vehicles and also has applications in driver assistance systems and road sign maintenance. Here an algorithm is presented using the Xilinx Zynq-7020 chip on a Zedboard to scan 1920×1080 images taken by an ON Semiconductor VITA-2000 sensor attached via the FMC slot. The PL section of the Zynq is used to perform essential image pre-processing functions and color based filtering of the image. Software classifies the shapes in the filtered image, and OpenCV's template matching function is used to identify the signs from a database of UK road signs. The system was designed in six weeks, and can process one frame in approximately 5 seconds. This is a promising start for a real-time System on Chip based approach to the problem of road sign recognition and also for using the Zynq platform for rapid deployment of these types of applications.

This project represents a first step towards bridging the gap between HCI and cognition research. Using functional near-infrared spectroscopy (fNIRS), we introduce tech-niques to non-invasively measure a range of cognitive workload states that have implications to HCI research, most directly usability testing. We present a set of usability experiments that illustrates how fNIRS brain measurement provides information about the cognitive demands placed on computer users by different interface designs.