eSchool Media

Unconscious AI in higher education

By Keith Rajecki
vice president, global public sector, education and research industry solutions , Oracle
April 12th, 2019

Artificial intelligence (AI), machine learning (ML), autonomous systems, robotic process automation, chat bots, augmented and mixed reality and many other buzzwords are flying around watercoolers and leadership team meetings across enterprises. It signifies the interest and the potential benefits to the organizations or institutions (in the case of higher education) and how these technologies can be adopted successfully to gain an advantage in the already very competitive higher education business. Part of AI is what is called unconscious AI. What does this really mean and what are the different perspectives of unconscious AI?

Different AI approaches

To explore unconscious AI, we first must understand what AI is and what different approaches are taken by technology providers and consumers to make AI effective and useful in daily life. While there are many different definitions and explanations about artificial intelligence, it is broadly regarded as the capability of a machine to imitate human intelligence. When evolving AI through neural networks, expert systems, robotic automations, autonomous systems and applications and many other technical tooling, we can observe three main types of AI and ML technology options adopted by the modern enterprise.

  1. AI point solutions serving a specific purpose (autonomous cars, policy automation)
  2. Using developer interfaces to provide AI capability (e.g. conversational interfaces, chatbots)
  3. Building AI capabilities into a technology platform as a service to be used by a wide variety of applications and services (AI platforms as a service providing a wide variety of intelligence across multiple applications and systems)
  4. Building AI capability into business applications to give a superior service to the end users (inherent AI embedded in business applications by design)

AI in higher education

There are many use cases and successful case studies of AI usage in higher education. In fact, higher education is an early adopter industry of AI for research and business with a lot of innovations coming from the sector. Classic uses of AI in higher education ranges from:

  1. Scholarship and financial eligibility selection: selecting the appropriate students for financial aid based on the institutional or government policy, demographics and other selection criteria using AI based policy automation systems would ensure consistency and policy alignment without any bias or prejudice.
  2. Advisement based on the personal needs and aspirations: with the growing needs and expectations of the modern student with entrepreneurial aspirations, AI could assist with the complex advisement requirements to suit the personal learning needs of each individual student.
  3. Marking and grading: AI based marking and grading systems will focus on content and objectivity devoid of the expression, language, construction, syntax and other biases that could affect the grading of work submitted.
  4. Institutional business applications: AI can play an effective role in institutional business applications to power the modern workforce with conversational interfaces and insightful analytics to drive efficiencies and provide a more satisfying employee experience.
  5. Academic recruitment and progression: with complex teaching, learning and research, student success and many other inputs defining academic tenure and progression, AI could play a significant part in academic recruitment and progression.

Whilst this is not an exhaustive list, these areas should provide some key use cases to illustrate AI and begin exploring unconscious AI in higher education.

Cognitive and unconscious behaviors

The human brain adapts and trains itself to its surroundings and is very capable of making decisions even though some of the decisions are unconscious responses. However, the inherent unconscious biases that exist within human behavior can also similarly affect human decision making. It has been noted (Anderson and Anderson 2007, Machine Ethics) that in self-learning AI systems these biases can seep through to the AI systems. Especially in conversational AI interfaces, these unconscious effects can get deep-rooted and skew the decisions making.

Removing unconscious bias

While AI is somewhat susceptible to unconscious biases, the AI can be trained and programmed to remove biases effectively. In the technology approaches above, the AI designers can make a concerted effort by using a broader cohort of different demographics to train AI models so that the AI decision making is unbiased and objective. This can be very effective in broader conversational interfaces, providing student interactions such as virtual advisement assistants or student service chatbots. The AI’s ability to scan a wider variety of data from a multitude of sources make this even more compelling and attractive to the users of such services.

However, in the case of financial aid and scholarship eligibility the reverse could be true. If there is a heavier bias towards equity than merit through AI selection and policy application, the objectivity and fairness is lost again. In these systems, AI should be trained and implemented to remove unconscious bias and to give both merit and equity equal opportunity. These systems depend on the demographic data and will only thrive if these data are provided unhindered to the system to make right decisions.

With business applications and platform services, using a broader sample set and diverse pools of developers and testers from different backgrounds can drive any unconscious AI impacts to a minimum.

In academic recruitment where referrals and recommendations play a significant part in potential candidate selection, the selection criteria itself could be subjected to unconscious bias. AI systems can play a significant part in the recruitment process, from as early as forming the job role and description, selection criteria and where to advertise the roles, even before the formal candidate recruitment process begins. It is seen that many modern human capital management systems are now well equipped with AI extensions in workforce planning.

We are still at the early stages of using AI in business applications and higher education systems. With AI becoming more prevalent and getting widespread use, advanced AI systems will be able to detect and correct themselves proactively and remove any unconscious biases autonomously from the AI systems without any human interaction. Don’t be alarmed just yet!

Our 250+ Clients Include