New evidence that setting up mixed human-robot teams can optimise data reading and analysis.
There is no denying the fact that Big Data is upon us. Now however, the question surrounding this ever-increasing mass of available data has become a qualitative rather than a quantitative issue. Ensuring proper analysis, extracting the most relevant information and putting it together – these are increasingly the challenges that will face analysts and operators. However futuristic it might sound, a team of academics from Princeton University, the University of Hartford and the University of California at Santa Barbara is advocating setting up mixed human-robot teams in such fields as surveillance. Their paper, subtitled ‘Integrating Cognitive Modeling with Engineering Design,’, seeks to demonstrate how robot support can help to reduce the errors made by a human operator in this field.
Mutual assistance to overcome intrinsic deficiencies
This is not some Asimov-like work of fiction, where every human operator is provided with a humanoid robot assistant. What this concept is all about is creating intelligent interactive systems for the purpose of collecting data required to carry out various different tasks. Firstly, at upstream level, sophisticated contextualising algorithms guide the robot assistant to feed back only information deemed to be essential to the operator’s decision-making process. Secondly, downstream, the system is designed to compensate for the operator’s human, physical imperfections, integrating 1) a procedure whereby the robot assistant monitors the human operator, e.g. spotting pupil dilation as a sign of fatigue, and 2) a security protocol which vetoes any decision – whether voluntary or involuntary – deemed erroneous. The academic paper posits an imaginary surveillance situation where, after receiving the information, the operator is given a ‘two-alternative choice task’ and demonstrates mathematically how a decision-making structure that is more efficient for a human operator can be implemented by making best use of a robot assistant.
Small (data) is beautiful?
Based on this scenario, the authors were able to quantify precisely the benefits of both upstream filtering of information and subsequent vetting of the decision by an automated system. To be clear, this is not about replacing the human operator, who is by definition the only one that can make a ‘human’ decision. The idea is to be able to provide him/her with ‘smart’ rather than ‘big’ data. In fact in this modelled scenario, an operator tasked to supervise a number of regions was able to concentrate greater attention on the regions judged most at risk, and saw his/her response capability clearly improve. Although the study is based on a fictional situation, with only two possible choices, and only one choice at a time, it will undoubtedly contribute some mathematical rigour to existing literature on this subject. Meanwhile, though it seems perfectly logical that technology will help to improve decision-making, a question mark still hangs over how successfully such technological aids can interact with human beings.