The promise of artificial intelligence and machine learning for higher education

Artificial intelligence, machine learning, and big data have afforded tremendous improvements to almost every field, including higher education. For example:

• The University of Aberystwyth in the UK has already developed—over a decade ago, in fact—the necessary robotic infrastructure to carry out scientific research on its own: developing hypotheses, conducting experiments, and analyzing required datasets. This represents a significant development in the experimental and research arena in order to ensure the accuracy of results and allow the human employees to focus on a more critical function
• The publisher Elsevier is using artificial intelligence to analyze literature reviews, measure plagiarism, and identify forged numerical or statistical features and details. This will ensure that unethical behavior is flagged before any publication goes live. Similarly, higher education institutions could benefit from such practices by implementing AI-induced mechanisms that would prevent malpractices in the assessment process, resulting in higher quality results.
• Intelligent chatbots based on natural language programming (NLP) are being already used by universities across the globe. The Technical University of Berlin (TUB), for example, has developed a chatbot system that can guide students around campus and help them choose their courses. Administrators at Spain’s University of Murcia were surprised to learn that its chatbot system answered 91% of 38,708 questions accurately. The chatbot enabled the university to operate outside of working hours and had a positive psychological impact on students—they became more motivated to use the chatbot over time, knowing there would be a tool to communicate directly with the university administration on an ongoing basis.
• Virtual assistants play a key role at many universities. Carnegie Mellon University, through its Open Learning Initiative (OLI), has developed AI-induced cognitive tutors to engage students. This had positive results in students’ overall performance and dedication levels. Similarly, at Georgia Institute of Technology, a virtual teaching assistant (TA), using IBM’s Watson Platform, was implemented in order to provide responses to about 40,000 questions during the course ‘Knowledge-Based Artificial Intelligence’. This ensured the prevention of low student retention rates and positive class engagement.

What Artificial Intelligence Can Bring to Higher Ed

Among some of the benefits of using information technologies in higher education, machine learning improves the learning experience and the ability to analyze the management of campus at all levels and better organize tasks. In addition to this, it allows the ability to receive new opinions from the input of a computer. Consider the following use cases.

A New Way to Plan Programs

Imagine planning only one academic program: How many variables should you be considering? How many options do you have available? How many combinations of courses, rooms, or students should you consider? How long does it take humans to reach a decision?

If we consider curriculum development, AI’s speed, accuracy, and consistency can ensure that an adequate subject selection and distribution will be established based on pre-set parameters by the educator or administrator. This would enhance the institution’s dynamic adaptation to the growing number of students and new programs.

Bias-Free Admissions Management

Machine learning and big data analysis fully exploit the power of artificial intelligence to expand the number of options and scenarios of any complex planning in our institution, such as admissions management.

Let’s take the example of Kira Talent, a Canada-based start-up that sells a cloud-based admissions planning platform. The company was able to shortlist up to nine different types of human bias during the student admission process, such as race, religion, and gender. However, the most critical and determining bias originates in the reviewer’s psychological exhaustion and the ever-growing variety of interviewers. This creates an inconsistency in the human-driven interview process which can be prevented by using AI. Alongside this, AI could also help to increase the accuracy of background checks, avoiding admission scandals like the ones that occurred in recent times.

Comparably, Taylor University in Indiana uses Salesforce AI-driven software that includes Protected Fields, which is a feature that displays pop-up alerts in order to avoid biases such as surnames that might indicate the place of origin, race, or even religion.

Large-Scale Learning Analysis

At the World Economic Forum in Davos, Switzerland, Dr. Katharina Hauck, senior lecturer in health economics at Imperial College London, talked about the future of AI and how it is beginning to enhance large-scale analysis, for example, in the health industry. It uses variable selection models through which it tests the importance of each factor with respect to the rest in scenarios where there may be more than 100,000 sub-models, allowing us to reduce estimates from weeks to a few days. This discipline could play a key role in the area of learning analysis, not only in curriculum quality but in the creation of more adaptive learning systems.

The long-term benefits of implementing learning analysis in higher education can be:

• Improving student retention. For example, at the University of New England, the student attrition was reduced to 12% and the students displayed a growing sense of belonging to the class and learning community in general.
• Supporting informed decision-making. In this case, at the University of Adelaide, educators were able to enhance the design of collaborative activities based on the data collected. Additionally, learning analysis can provide inputs about the most suitable teaching assistants to be assigned to a particular group of students, as it is applied at the University of Edinburgh.

Scenario Planning and Improved Decision Making

Machine learning information allows administrators to explore possible future scenarios simulating realities at low cost without incurring many of the risks of real experimentation. With the help of human questions and a good conceptual framework, smart machines can help managers to review vast volumes of data to discover patterns.

In addition to that, the use of machine learning in management can show the long-term consequences of certain short-term decisions. Thus, machines can help identify unexpected consequences of a resolution or discover value niches with rapid experimentation, at high speed.

For example, artificial intelligence is used to track student performance as it occurs at Georgia State and Arizona State where this technology is being used to predict scores and assure that deserving students reach their full potential and prevent those who are underperforming from dropping out. Universities will need to cope at the pace of technological development which will require the creation of new jobs, departments, and degree courses.

In this way, higher education executives can consider the various scenarios available in preventing moral, ethical, or cultural disturbances from taking place.

Implementing the Technology: Recommendations

Higher education institutions utilize Enterprise Resource Planning (ERP) systems which allow us to manage all the influx of information. However, they fail to automate the delivery of solutions. Thus, many universities have realized that an ERP isn’t adequate to handle so much workload because it cannot dynamically manage the automation of the academic planning process.

It is here where machine learning techniques become essential using what is termed as “intelligent decision-making systems” that allow the elimination of several inconveniences of ERP and other traditional decision-making systems, such as lack of historical data references and lack of dynamic functioning. AI will be able to effectively manage simulations and predictions in areas such as decentralization of campus management, general planning, student profiling, and collaborative work.

The best way toward a full-fledged implementation is to provide an adjustment time frame period crossing the required stages upon this technological revolution:

• Implementation: the institution must define its own technological requirements depending on the work methodology, structure, program distribution and financial investment required.
• Rejection: confusion and distress will take place as it is required to address the fact that these new technologies will take away several job roles and create unemployment at a certain rate. Data privacy will also be an active concern as data requirements will increase with the development of technology and the need for a constant flow of real-time dynamic data where data ownership might become an obsolete concept. Similarly, the staff will be undergoing a fundamental upskilling process when it comes to new teaching methodologies with the utilization of various tools and, consequently, reluctance toward adaption will be present.
• Adjustment: upon the regular use of this technology, the staff and students will understand the need for implementation and will show interest in understanding the mechanisms involved and long-term impact on institutional productivity.
• Acceptance: also known as the Technology Acceptance Model (TAM), which explains the acceptance of new technologies by users based on three key elements: the technological usefulness, easiness to operate, and attitudinal approach.
• Continuous Development: after accepting the need and interactivity of this technology, users will feel the need to take part in future advances that these systems can bring about, acknowledging the durable benefits provided.

By Dr Raul V. Rodriguez, Dean, Woxsen School of Business, Woxsen University