By Marcelo Cabrol

Back in 1984, educational psychologist Benjamin Bloom hinted at what promised to be a revolution in education. He reported on data showing that students working with individual tutors reach achievement levels as much as two standard deviations higher than students in conventional instruction (50% of tutored students score higher than 98% of the comparison group). This meant that individual tutoring had the potential to transform education. However, there was one important catch: one-on-one tutoring is the most expensive form of instruction. The economics of individual tutoring is at the core of Bloom’s 2 sigma problem, and it has not yet been resolved, until now.

Almost forty years since Bloom’s formulation, the era of generative AI is here. The capacity of AI to create new content from training data has been built up for several years, but 2023 will be remembered as the year of Generative AI. Excitement over this technology is palpable, early pilots are compelling, and its transformational potential in numerous fields is enormous.  

Take the world of employment as an example: generative AI has the potential to change the anatomy of work, augmenting the capabilities of individual workers by automating some of their activities. According to a recent McKinsey study, current generative AI and other technologies can potentially automate work activities that absorb 60 to 70 percent of employees’ time today. Workers faced with this situation will need support to learn new skills and revamp existing ones, and some will change to existing or entirely new occupations.

A radical change in education is needed, and generative AI can make it possible. 

Generative AI can provide intelligent individualized tutoring. Customizing feedback and support to the needs of each student is at the core of Bloom’s formulation.  Both established and new companies are working in this area.  For example, Khan Academy and DuoLingo are piloting GPT-4 powered tutors trained on their unique datasets. 

Tutoring and personalized learning are two sides of the same coin. Generative artificial intelligence can analyze students’ learning patterns and personalize content and teaching methods accordingly. This can help students learn at their own pace and in a way that suits their individual learning style. 

Furthermore, teachers can become tutors by personalizing their practice. Generative AI can help teachers in at least two ways: improving the quality of their assessments in both accuracy and efficiency and enhancing content creation. Going well beyond automated grading, AI-based feedback systems can offer constructive critiques on student writing, including feedback aligned to different assessments, which helps students elevate the quality of their work and fine-tune their writing skills. AI-powered tools can also help educators create better materials for their students. For example, AI-powered tools can help identify knowledge gaps in existing materials, suggest improvements or even generate new materials entirely. 

Overall, generative artificial intelligence has the potential to revolutionize education by making it more personalized, efficient, and engaging for students. However, like with any other revolutionary technology, all these benefits come with a “warning” label. The following are the risks and ethical considerations that we need to know and manage.

First, there are issues that underlay the technology itself. AI algorithms can be biased if they are trained on data that reflects existing prejudices in society (Leonardo Nicolletti and Dina Bass’ study in Bloomberg is very insightful). AI-powered educational tools could reinforce existing inequalities and discriminate against certain groups of students. These tools may also collect sensitive student data, such as their learning preferences and performance data. This data must be handled carefully to protect student privacy and prevent misuse.

Other warnings come from the negative dynamics that AI and personalization can create. One is the lack of human interaction: while AI-powered educational tools can provide personalized feedback and support, they cannot replace the human interaction that is important for social and emotional development. An over-reliance on AI could lead students to isolate themselves and lose important social skills. Associated with the loss of social skills is an extreme reliance on technology in which students become too dependent on technology and cannot learn without it. This could lead to students lacking critical thinking skills and being unable to solve problems in the absence of technology.

The good news is that we know how to mitigate these risks. The basic methods to mitigate the threats related to Generative AI in education are the following:

Diverse data collection: The data used to train AI models should be diverse and representative of the entire student population. This can help reduce the risk of AI perpetuating existing biases.

Regular testing for bias: Regular testing for bias can help identify and address any prejudice in AI-powered educational tools. This may include testing for bias in AI-powered recommendations, ratings, and feedback.

Human oversight: It is important to have human oversight to ensure that decisions made by AI are fair and unbiased. This can include teachers reviewing AI-generated recommendations and feedback before sharing them with students.

Transparency: AI-powered educational tools need to be transparent in their decision-making processes. This can help build trust between students, teachers, and parents and allow them to understand how decisions are made.

Continuous improvement: AI-powered educational tools must be continually improved based on feedback from students, teachers, and parents. This can help ensure that AI is used to benefit everyone and not perpetuate existing biases.

At the IDB Group, we are committed to promoting the responsible use of new technologies. We have a generative AI working group led by our Technology department and the emerging technologies laboratory, TechLab. It is a multidisciplinary team that brings together experts from different departments and regions to generate a positive impact of technology in Latin America and the Caribbean. For more details on their initiatives and progress in this area, we invite you to consult the Tech Report on Generative AI.

In parallel, at BID Lab, our innovation lab, we work with initiatives such as fAIr LAC, the largest alliance in our region for the ethical and responsible use of technology, to realize the promises of AI in education and beyond, since only by applying ethical lenses can we use technologies to build fairer and more prosperous societies.

By properly balancing the benefits and risks of generative AI, Bloom's problem can now become Bloom's promise. AI opens the door to a future where teaching and learning are more effective, equitable, and deeply personalized, allowing empowered students to reach their full potential and teachers to focus on teaching and fostering meaningful connections with their students.