Lifelong Unlearning

I was talking to a colleague the other day and we were discussing how our formal systems always assume that learning is additive. We are always talking about more courses, more skills, more knowledge. But in practice, growth also comes from subtracting. Anyone who has worked hard to drop a bad habit knows how tough it is to let go. Whether it is a habit, our assumptions and biases, or even poor work practices and ways of working, "unlearning" is hard to do but has nothing to show. There is no concrete evidence of unlearning or any artifacts or portfolio to show what and how much we have unlearned. We don't give out certificates for what people have stopped doing!

Unlearning is subtle, but it is powerful. After unlearning my fair share, I know that it has made space for better judgment, greater adaptability, and more thoughtful practice.


I wrote about this previously, in 2012, in my blog post titled:
Emptying Your Cup: Unlearning to Learn.

The point I am making is that not all growth comes from learning something new. Sometimes, it comes from letting something go. And we have to be intentional about it.

What did you let go of recently that contributed to your growth? How do you encourage others to unlearn? How do we design "learning" systems that also enable and support unlearning?

Thoughts?

----------

Additional reading:

• Chapter 15: Learning, relearning and unlearning
  
  November 2024 Open Books and Proceedings
  DOI:10.38140/obp1-2024-15
  License CC BY 4.0
  Authors:
  Rosemary Akinyi Menya-Olendo
  Lucy Mawang
  Kenyatta University

https://www.researchgate.net/publication/387466020_Chapter_15_Learning_relearning_and_unlearning

In this chapter, the authors describe the following key terms:

Learning is contextualised as the continuous ability to acquire competencies relevant for the 21st century. This form of learning enables people to continuously improve their performance, expand their horizons, and equip themselves for the future. It further allows individuals to grow and develop personally and professionally throughout their lifespan. Turk (2023) observes that this type of learning occurs in various contexts, such as formal education, informal learning, and experiential learning. 

Unlearning means leaving behind old, outdated, and obsolete knowledge that is deemed inefficient in ad-dressing current challenges. It therefore entails questioning one’s assumptions and beliefs, and opening up to new perspectives that can help solve present problems. No wonder Turk (2023) postulates that the unlearning process may be challenging, as it necessitates confronting personal biases and preconceptions.

Relearning is the process of learning something again, often in a new or different way. It involves building on previous knowledge and experiences to gain a deeper understanding of a subject or skill. Relearning is important because it allows individuals to update their knowledge and skills in response to new information and changing circumstances (Turk, 2023).

• It turns out that unlearning isn't just a human challenge. AI and LLMs are wrestling with the same problem. A paper submitted to ICLR 2026, Lifelong Unlearning for Multimodal Large Language Models (Li et al.), tackles the problem of teaching AI systems to selectively forget outdated or problematic information. 

"Multimodal large language models (MLLMs) are trained on massive multimodal data, making data unlearning increasingly important as data owners may request the removal of specific content. In practice, these requests often arrive sequentially over time, creating the problem of MLLM Lifelong Unlearning."