In the research population, Most researchers typically spend about 5 hours per session of a study, 1 hour to prepare the study, 2 hours during the study, and 1 more hour managing the data afterwards. For any study of size 50-200 people, this easily spans to 200-800 hours. Moreover, currently, unpaid and overworked assistants must spend inordinate resources to collect this data. This results in human errors that waste funding, lost time that could be spent on more creative research tasks, and small datasets that limit generalizability to the larger population.
Not long ago, statistical/visualization software, word processors, and communication tools advanced science by standardizing processes, reducing errors, increasing speed, and fostering more creative pursuits. Data collection is also becoming automated in the health sciences where the study is not on humans. Ubiquitous computing can be the next step, by doing the same by collecting a large amount of data from human subjects in social science, health science, and HCI labs.
Recent applications of computer vision to detect human actions and expressions coupled with natural language processing would enable scientists to let the lab room itself guide participants through simple experiments and studies. Imagine a participant entering a “smart” lab room by activating an electronic lock, conversing naturally with an AI during the study, and being compensated afterwards. No humans needed.
I used Haar Cascades, a Computer Vision deep learning algorithm, to recognize a human face and invite the participant to sit, followed by video directions for the study.
- Seamlessly implement Computer Vision with a voice assistant for study directions, and NLP to understand requests made by the participant ("I would like to talk to the researcher").
- Make the project open source and encourage open source development of features among the research community so that it can be used in diverse situations and implementations worldwide.
Integrating CV, NLP, and a Voice assistant together and making it one seamless AI would be a challenging project.