Goal Alignment

A skill that opens up new potential for educators, and perhaps, humanity itself

Educational practices have been informed over the years by many fields of research. Psychology, philosophy, neuroscience, cognitive science and more have played a role, and to good effect. One field of research that has not yet had an impact on educational practice is machine learning. This is the field of study that has produced all the AI wonders that you see emerging every day. And while it is quite clear that AI itself will reshape education, the machine learning methods behind it have so far not been a source of much inspiration. But we think that could change and this change could bring new possibilities to education.

---

Have you listened to The Optimalist Podcast?

At The Optimalist, we are exploring the optimal way to educate. Our podcast takes on that challenge through conversations with educators, psychologists, doctors, technologists, and innovators. We believe this is the way to face the unique issues we now face as we enter the age of AI.

This week, host, Sara speaks with Steve Tetreault, a former teacher turned passionate school librarian. Throughout the conversation, Steve emphasizes the vital role school libraries play in authentic learning and his belief in librarians as champions of new technologies like artificial intelligence (AI) in education. He highlights the socialization and community-building aspects of libraries as well as the importance of information literacy skills development.

Lastly, the episode delves into the power of fiction in fostering empathy and communication, the evolving role of school libraries in the digital age, and the potential of podcasts to create modern reading communities.

You can listen to a new episode every Wednesday, right here in our newsletter, and anywhere you can subscribe to podcasts. Click below to listen to Sara’s conversation with Steve.

---

Machine learning has more typically been the receiver of insights from education, psychology, cognitive science, etc. It’s a field of research that dates all the way back to the 1950’s when the first computers were made. While the field was held back for years by computation limitations, it was mostly an ignorable field for educators. It was a technical pursuit without any clear applications. But now that things have started to work - and they have really started to work - insights can start to flow in reverse.

Our thinking is that perhaps where big breakthroughs occur in machine learning - particularly in areas that have been hard to prove and agree upon in human development - it can bring some fresh ideas to the table. Of course, it’s worth recognizing upfront that it will be an imperfect source of insights. There are clear and meaningful differences between human biology and computers. Not everything will translate. 

We started to think this way about it a few years ago around the release of Open AI’s GPT-2. Our interest in attention was partly inspired by the breakthrough research paper Attention is All You Need that set the computer architecture foundations for AI. We previously wrote about how creativity is evolving, how AI has provided new insight into what creativity is, and how this insight might soothe the debate about how to develop it. And we’re currently digging into how gradient descent can be used to inform educational practices. 

All of these discoveries are the result of studying the implications of machine learning. 

And for today’s post, we want to bring you into what we’re currently integrating into our work from ML: what is known as the alignment problem. 

Alignment is a word commonly used in AI safety circles. It speaks to the challenge of training AI to align with human values - so that we flourish as a species and avoid the myriad of doomsday scenarios that have been imagined. It’s most often discussed as an engineering challenge - and all the major AI platforms have dedicated safety teams actively trying to figure it out. It most often uses a machine learning method called Reinforcement Learning from Human Feedback (RLHF). Alignment is regarded as a really hard problem to solve. This is because what makes AI alignment so difficult is that we are not well aligned with each other, or ourselves. 

We are self-destructive and do things that we regret. We often do things that are at cross purposes to our own goals. We cause harm to others, knowingly and unknowingly - at small and large scales. When presented with a powerful tool like AI, this can magnify all those misalignments. And most doomsday scenarios are really imagined from this fact. 

All this being said, the alignment problem is rarely discussed as an educational one, and we think this is shortsighted. 

What if it’s possible to be more aligned, with ourselves and with others? What if it is a skill that can be taught, measured, and perhaps produced as an outcome of education? 

We think it’s possible. 

Reflection is the original act of self-alignment. A good reflection improves your thinking about learning and goals, and helps you build the self-awareness and self-regulation to hold yourself accountable. And most educators already regard reflection as a necessary virtue and practice.

Educators also likely already understand the idea of goal alignment in practice quite well. It’s when student actions and learnings reflect a stated lesson objective. It’s when they find common goals and collaborate well together. It’s when they find motivating goals within their work and hold themselves accountable to follow through on them. 

The obvious reason why goal alignment is so hard overall is that not all goals align. Some goals are completely incompatible with others. There is no way to meet or satisfy one without failing the other. This will always remain true and it would be foolish to think otherwise. This is why we need the engineering approach to AI safety - learning through human feedback (RLHF). But relying solely on this path to solve the problem is also risky - it could threaten rights, introduce onerous censorship, and more. 

To avoid this, we need to be better than we’ve ever been at this.

The way we need to be better is by committing ourselves to recognizing first the role self-alignment plays in this. To be better at aligning goals, we must be better at recognizing them within ourselves - understanding they are fluid and flexible. That setting a goal doesn’t mean sealing your fate. And that kind of skill - to be bendable, moldable, changeable - comes with motivation to be better. A byproduct of self-reflection.

Because the reality is that there are almost always more options for goals than we think. You just need to have the skill to uncover them. There are many more win-win scenarios than we typically realize in life, at school, or in employment. Perhaps one way to think about it could be that we need more goal creativity - to express more goals and to be less fixed and more fluid in our thinking. We need to be more process-oriented over more outcome oriented. This is something we can all get better at. 

And it’s something educators could help with. 

We recognize that as we say that, that’s adding one more thing to an educator's plate. This is the dealbreaker when it comes to most new ideas in education these days. So instead of asking for one more thing, goal alignment is one of the reflective insights we’ll be automating with AI in Mirror software. You can think of it just like Reinforcement Learning by Machine Feedback (RLMF). Regular reflection will help students see their goals as part of a larger connected web, instead of isolated and perhaps limited in scope.

The bottom line is, we think it’s so important that we want to find a way for you to practice it and prioritize it consistently without it becoming a burden. In fact - we want you to wonder how you’d been operating for so long without it before. 

Without much added effort, educators will be able to assess goal alignment, and set up customized reflections that accelerate the development of this important skill. This will bring immediate and practical benefits to educators too. They can assess if students understand their objectives. It will help students develop better goals and stronger self motivation. It will help foster better teamwork and collaboration. Overall, it can help create a more harmonious learning environment that everyone is seeking. 

Our hope is that if we can measure it, we can help make it matter as a skill - the first specific skill we’re proposing for inclusion in Even Higher Order Thinking Skills (EHOTS). And if we can make it matter, that means educators will be contributing to this important societal challenge and to their students’ future thriving. 

What do you think?

---

🤖 Adaptability in Education Leadership Panel Series

In many of our conversations with school leaders, a common sentiment emerged: not enough support exists for school leaders when it comes to helping their communities of educators navigate change. The question: How can we help teachers navigate complexities successfully without making sure we’re confident doing so ourselves?

The Adaptability in Edu Leadership Panels are your passport to being a part of a world where change isn't just embraced, it's mastered. Imagine being part of a community where the best minds come together, sharing strategies and experiences that help light the path forward. That’s what we’ve been building.

These panels are about actionable insights. You'll gain the tools, motivation, and skills to navigate complexity with confidence. We will help you reshape how leaders conquer challenges and harness change in their schools - as models of it themselves. Also… it's free!

If you are a leader who would like to participate in an upcoming panel, please email sara@swivl.com.

🤖 Introducing Mirror by Swivl

Mirror is a new a tool that uses AI to automate reflection in order to align classroom goals and develop even higher order skills.

The act of reflection possesses an untapped potential that can unlock new levels of human growth and development if practiced at a deeper and more regular pace. At Swivl, we build tools to realize this potential in schools… and beyond.

Mirror is where students (and you!) will meet their potential self.

We are looking for volunteers in schools throughout the US and Canada to trial Mirror in their school right now for a 30-day period. Click here to fill out a simple form to request a demo device, and someone will be in touch with you very soon.

---