Supporting Your Team with Machine Learning
The training and recruitment of staff is a complex process made all the more difficult by the fact that as humans, we nurture a set of thoughts and feelings about what we think makes a good employee that isn't always based on evidence. Using any behavioural data that we have on our existing team may allow us, through a participatory process, to identify not only how an employee may play to their strengths in the workplace, but how best to approach a mentor-mentee relationship, how to shift between optimised work behaviours, and in some cases, it can help change the way the organisation sees and talks about itself.
This paper explores the relationship between unsupervised machine learning models, and the mental models of those who develop or use them. In particular, we consider unsupervised models, as well as the 'organisational co-learning process' that creates them, as learning affordances. The co-learning process involves inputs originating both from the human participants’ shared semantics, as well as from the data. By combining these, the process as well as the resulting computational models afford a newly shaped mental model, which is potentially more resistant to the biases of human mental models. We illustrate this organisational co-learning process with a case study involving unsupervised modelling via commonly used methods such as dimension reduction and clustering. Our case study describes how a trading and training company engaged in the co-learning process, and how its mental models of trading behaviour were shaped (and afforded) by the resulting unsupervised machine learning model. The paper argues that this kind of co-learning process can play a significant role in human learning, by shaping and safeguarding participants’ mental models, precisely because the models are unsupervised, and thus potentially lead to learning from unexpected or inexplicit patterns.
Learners’ mental models, unsupervised machine learning, co-learning process, training, recruitment
About the Authors:
Dr Carmel Kent
Carmel is the head of data science at EDUCATE and a senior research fellow at UCL. She is a computational social scientist, with a research focus on Artificial Intelligence for education and learning analytics.
Her other areas of interest focus on online learning communities, interactivity in online discussions and collaborative learning technologies. She has 20 years of industry and academic experience, having worked as a software engineer, data scientist, entrepreneur, teacher and researcher, and for IBM research, EdTech and healthcare providers’ companies and a number of startups.
Dr Mutlu Cukurova
Mutlu is an academic faculty member at University College London and has a particular interest in researching the potential of emerging Educational Technologies such as Artificial Intelligence and Learning Analytics to continuously evaluate and support human development. In addition to this, Mutlu works with UNESCO’s international expert group on ICT in Education. He is Director of Research at EDUCATE and sits in the working group of UCL’s Grand Challenges on Transformative Technologies.
Mutlu's work is interdisciplinary and encompasses research in learning sciences, psychology, computer science, and Human-Computer Interaction. For more, visit his profile here.
An AI enthusiast who is always on the lookout to learn about the latest AI algorithms and the ‘magic’ behind them, Ali is a postgraduate researcher at UCL Knowledge Lab with a research focus on exploring transparency through frequentist and bayesian statistical techniques. He also explores and compares the learning trajectories of reinforcement learning agents in simulated environments with human learning in the real world.
A firm believer of education’s role in changing the world for the better, Ali runs a number of charity schools in Pakistan.