Online learning giant Coursera launched brand-new AI-powered learning tools at their flagship conference last week, as part of a wave of institutions and corporations seeking to revolutionize education with generative artificial intelligence. Chief Content Officer Marni Baker Stein talks to us about the potential of this technology in the education sector, and the challenges that students and educators will come across as it changes the face of learning. She says:
The first is how do we take all this amazing content and expertise coming out of universities and more efficiently and effectively create highly interactive high-performing online courses and do that at a standard of fidelity that's even and consistent across the board without, without having to create armies of staffing to support that effort?
And then the second problem we're trying to solve is, how do we introduce Online Learning at the scale of the world to provide sort of ubiquitous, personalized contextualized tutoring or coaching for learners as they're moving through learning experience?
However, some of the challenges of using generative AI in the education sector, as outlined below, highlight how companies such as Coursera will have choppy waters to navigate over the coming months and years.
1. Defending academic integrity
Since the widespread media coverage of the likes of ChatGPT, institutions and leaders across the globe have raised concerns that generative AI threatens their assessment processes. This concern is borne out of the ability of generative AI to Pen entire essays and reviews from a couple of specific, user-defined queries. Universities are already responding, with one essay-based assessment being removed from a Computer Science module at University College London and working groups being established in faculties across the UK and abroad.
Coursera’s use of AI is slightly different, however. Its launch of ‘Coursera Coach’ will act instead as a learning mentor, where users can design bespoke courses and get instant personalized feedback on their work. Baker Stein adds:
You can literally type in a learning objective like this course is about x. It has five learning objectives which are ABCDEFG, and the application will scour Coursera’s catalog and bring up the most relevant atomic content to those learning objectives.
Baker Stein agrees that there is work to be done to reassure educators and students that developers are taking academic integrity seriously, and noted that the risks are very much still present and the work to mitigate damage is ongoing. She adds:
Many of the assessment types and approaches are now going to be very difficult to protect from an academic integrity point of view.
We're embracing generative AI but we're taking academic integrity very seriously. And I think it's going to drive a transformational moment in the design of learning experiences.
The overwhelming tone here is optimistic. Assessments, as we have come to know them, will need to change as AI plays an increasing role in our education system, but Coursera notes that this need not be a bad thing and that these tools could prove transformative for assessment in an ageing industry.
2. Data Protection in the Age of Generative AI
We were keen to hear how Coursera seeks to tackle a growing concern amongst AI sceptics, how the vast swathes of personal data that this technology requires to function would be protected in a world where data is a valuable commodity.
Generative AI rests on the speedy application of complex algorithms to specific user data. Machine Learning programs are also ‘trained’ on this sensitive user information. Firms as recently as this year have already been ordered by the FTC to destroy algorithms that were built on illegally obtained or illegally stored data. The matter of data privacy is therefore a point of contention in debates surrounding AI, and this is relevant in the education sector too. We ask Baker Stein about how Coursera is meeting these concerns and how they seek to protect learner’s data whilst still providing highly personalized experiences. She says:
One of the ways that we're dealing with it is we are working with Microsoft Azure on the process, we've adopted their principles, we've established our own principles for ethical use of technology and we're working through their infrastructure.
Coursera’s launches are embedded in the Microsoft Azure data management platform, a service that automatically encrypts data and stores it in its 600 global data centers. Coursera states that the way it collects and stores its data is also designed so as to as to deter its corruption or misuse. Baker Stein describes the user data collected by Coursera Coach as “stateless”. The responses given by the tool are contextualised to the specific questions being asked by the user.
Amongst this optimism however is an underlying uncertainty that has become characteristic of the AI revolution. Regulating the ethical use of user data is a challenge that governments around the world are still grappling with, and whilst companies like Coursera are claiming to address these, machine learning has changed gears. Data protection principles must develop rapidly to protect the rights of users worldwide. Some of the principles we have today, including those found within GDPR, will be applicable to the current AI models being created - but we will also need new principles. And as we know, the speed at which regulation is created is often at odds with the pace of technology development.
3. Ethical and Human-Scale Development
Underpinning all of this is the need to retain the human touch in education. Developers across the industry are confident that working alongside experienced educators will generate tools that are impactful and respectful. We asked Baker Stein what these conversations are looking like in the higher education institutions Coursera is partnering with, and she noted the equal parts “Caution and exuberance” as leaders grapple with this new technology. Positivity was the overwhelming outlook, and companies like Coursera say they are working with institutions to develop their products. Coursera argues that the development of AI educational tools is in conjunction with the institutions that become their largest clients.
It is hard not to get excited about the potential applications of AI in the education sector. The industry has been calling out for modernisation and democratization for years as it has been mired by antiquated assessments and a lack of personalisation. Launches such as Coursera Coach represent one such direction that the sector could take, but certainly isn’t the only one. Concerns may still be raised regarding access; concerns that are all too common as we toy with AI’s capabilities. Developers must maintain a constant awareness of scale and the industries they seek to revolutionize, lest they risk speeding ahead and leaving learners and educators behind.