Trust is the Silver Bullet
Generative AI is straining it, just when we need it most. Here's how to rebuild it.
Thank you for being here. Please consider supporting my work by sharing my writing with a friend or taking out a paid subscription.
There are many challenges facing academia today, but the breakdown of trust is at the core. We have entered the age of suspicion, and it’s only been accelerated by the diffusion of generative AI across nearly all aspects of academic life in teaching, learning, and scholarship. Students and their parents suspect that college may not deliver on its promise, professors suspect students of not doing the work, authors suspect reviewers of outsourcing their reviews to AI, students suspect administrators of just trying to keep the peace—on and on it goes. Everyone suspects everyone.
But even if you agree that this breakdown in trust is a big problem, fixing it can seem intractable. Once trust has been broken, can it be rebuilt?
I believe trust can and must be rebuilt. As we grapple with the upheaval of our current moment, I am convinced there are few things more worthwhile than working to rebuild trust in every domain of our work. Building trust is the closest thing we have to a silver bullet solution to the current challenges we face.
Trust is character plus integrity
But what exactly does it look like to build trust? First, we need to start with a diagnosis of the components of trust.
One helpful way to think of it that I’ve been spending some time with lately is presented by Stephen M. R. Covey—the son of Stephen R. Covey, author of the well-known book *The 7 Habits of Highly Effective People*—in his book *The Speed of Trust*.
In the *Speed of Trust*, Covey decomposes trust into two main elements: character and competence. Character is about who you are, whereas competence is about what you can do.
In his framework, Covey further breaks down the two main elements of character and competence into two main subelements. Character, he argues, is composed of integrity and intent. Do you do the right thing when no one is looking, and do you act in the best interest of the broader community in mind? Competence can similarly be decomposed into two pieces, capability and results. Capability is about whether you possess the skills to do a certain task well, while results describe your ability to get stuff done.
Trust is built on each of these four elements, and a deficiency in even one of them is enough to destroy trust. Some of these are more obvious than others. It’s pretty intuitive that if you don’t think someone is acting in the best interest of the group (intent), then you can’t trust them, even if they exhibit the other three qualities. In a similar way, a person’s lack of integrity (doing what’s right even when no one is looking) means that you can’t trust that person, even if you might believe that they have the best interests of the group in mind.
These two aspects of trust feel relatively easy to understand. But the other two factors that Covey identified have felt particularly helpful as I have been thinking about the breakdown of trust in schools. Although they’re a bit more subtle, they’re important pieces of the model to understand how trust can break down even when you’re dealing with people of strong character.
The importance of competence becomes clearer when you consider a situation in which you need to rely on someone to deliver something. Even if people have strong character, you cannot trust them if you don’t believe they have the capability to deliver on their responsibilities. Perhaps you don’t think they have the skills needed to complete the part of the group project that they’ve been assigned. No amount of integrity and good intentions can overcome this real or perceived lack of skill. Similarly, even if you have the capabilities needed, if you haven’t proven yourself able to deliver on what you promise, then you can’t be trusted. I’m sure all of us can recall at least a few group projects where we’ve had someone who couldn’t be trusted because they never completed their deliverables on time.
Generative AI is straining not only trust between, but also trust within
Generative AI is placing a significant strain on trust in our classrooms. Most of our conversation has been focused on thinking about academic integrity as a violation of character. In other words, a student cheats because of a weakness either of integrity or of intent. We succumb to the temptation to do something wrong when we think no one is looking or fail to consider the implications of a shortcut on our peers.
But Covey’s framework highlights another aspect where generative AI is making it difficult to build and maintain trust. Even if we believe our students to be people of strong character, understanding that capability is a critical piece of trust explains why strong character is not sufficient for trust.
One of the reasons that trust is so hard to maintain with generative AI is that the temptation to outsource the hard work of learning to generative AI is so seductive. Even if we attempt to use generative AI with the best of intentions to help us learn more effectively, it’s so easy to fall into the trap of over-reliance.
What’s worse is that the omnipresence of generative AI in our toolboxes can damage our self-trust. If we regularly turn to generative AI to help us with our work, the doubts may begin to creep in: Can I really conquer the blank page staring back at me? Can I come up with creative ideas without the help of a genAI brainstorming tool? When I get stuck, can I get unstuck without needing AI to break me out?
As we think about generative AI, we need to think about all four of these elements of trust and be mindful to strengthen them both with ourselves and with the people around us. We need to regularly examine our own integrity and intention. Are we really acting rightly even when no one is looking? Are we acting in the best interests of our colleagues and peers?
But beyond these elements of trust, we must also be thinking about how generative AI is reshaping the way we think about our competence. How is generative AI forming our own self-identity as we think about our skills and capabilities? Do we still feel confident that we can think hard and solve problems without our favorite AI tool in hand? Can we still deliver results even if the Anthropic API is down or our Claude token budget is depleted?
These are questions that all of us need to be asking, not just of our students but of ourselves. Only after we’ve asked these questions can we see the way forward.
Trust is like a tree
The metaphor Covey uses in his book to describe trust is a good, if simple one. He says that trust is like a tree. Integrity is the roots, intent is the trunk, capabilities are the branches, and results are the fruit.
Trust starts with integrity; if you don’t have it, people can’t trust you. But this is only the foundation. You must also have good intent. People must believe not only that you will do the right thing even when people aren’t looking, but that you have their best interests in mind too. If people doubt that you are acting in the best interest of the team, they can’t trust you.
But trust goes beyond that. Built on the trunk are branches, the capabilities we have to be able to get things done and carry out the tasks that we have taken on. These, in turn, connect to the results, the fruit. Each step matters, and a breakdown anywhere in the chain means there is no fruit.
As we think about generative AI in the classroom, Covey’s framework gives us not just a diagnosis of where trust is breaking down, but pointers on how to rebuild it. It starts with a firm foundation. If students are not engaging in their work with integrity and an understanding of how their intentions shape not only their learning experience but the experience of their classmates, then we’ve got serious problems. We need to emphasize the importance of doing the right thing and repairing integrity by forthrightly admitting when we have failed to uphold the standards we all aspire to meet. Likewise, we need to regularly revisit and critique our intentions. Why are we all here, and what does it look like for our actions to align with the flourishing of the whole learning community?
But once we have a strong foundation of character, we also need to address the challenges of competence head-on. We need to be straightforward with our students about the fact that AI does not democratize expertise, but amplifies it. To clearly outline how outsourcing your thinking to generative AI stifles your own intellectual development and stunts the development of your capabilities, capabilities that you will need to demonstrate not only for others to be able to trust you, but for you to be able to trust yourself.
Finally, we’ll need to talk about results. About how results matter, but the path of getting results without building capability leads to hollow competence. About how getting the product without going through the process is missing the entire point of what we are trying to do here. To engage in the nuanced conversation about using AI as a tool to amplify our work and create real value while simultaneously being aware of the ways that it may be leading us down a path of self-deception.
There are no silver bullets to solving the problems that AI is creating for education, but building trust through the components that Covey describes—integrity, intent, capabilities, and results—is about as close to one as we’re going to get.
Got a thought? Leave a comment below.
Reading Recommendations
Great post here from the Cosmos Institute, riffing on Humboldt’s idea of bildung for the modern challenges around education and AI.
Bildung is, at its normative core, anti-servility: the effort to form people who cannot be reduced to instruments of external authority, whether state, market, or algorithm.
The people who use AI well right now are drawing on judgment they formed before these tools became ambient. They know when to trust an LLM’s output and when to push back because they learned to read, argue, and sustain attention under conditions in which those acts were not so easily outsourced. They bring something the tool cannot supply.
Nietzsche thought secular liberals were living off the moral capital of a Christianity they had officially abandoned: inherited capacities that could persist for a time even after the culture that formed them had ceased to renew them. The kind of judgment this essay is defending may be a similar afterglow, formed in a world before AI mediated everything.
This was also very good from Minas Karamanis, “The machines are fine. I’m worried about us.”
We have centuries of accumulated pedagogical wisdom telling us that the attempt, including the failed attempt, is where the learning lives. And yet, somehow, when it comes to AI agents, we've collectively decided that maybe this time it's different. That maybe nodding at Claude's output is a substitute for doing the calculation yourself. It isn't. We knew that before LLMs existed. We seem to have forgotten it the moment they became convenient.
A few tweets worth reading too.
The Book Nook
While the ideas in The Speed of Trust may seem simple, I’ve enjoyed listening to the audiobook and getting some more insight into not only these four elements, but the many ways that Covey talks about the practices that help to support the ongoing development of trust.
This video from Praxis board member Mike Bontrager does a good job of illustrating Covey’s trust framework. The whole video is worth watching, but the part that is most relevant to this post starts at timestamp 2:05.
The Professor Is In
Had a blast serving as a judge for the Accelathon this weekend, hosted by the Claremont Accelerator. Fun to see what students were able to build with AI in just a few hours over the course of the day. Also great to hang out at The Hive with some friends like Josh Tatum from Storyhouse Ventures. Kudos to the CA team for putting on such a great event.
Just over a week from when early registration closes for our conference at Harvey Mudd next month. Check out more at our website and register before prices go up! Abstracts for poster, lightning talks, or discussion sessions are similarly due soon (April 22nd). Feel free to ping me with any questions.
Leisure Line
Had some fun baking this weekend. First up were some lemon poppyseed (actually chia seed) muffins that were very delicious. Second was a new chocolate chip cookie recipe. Both winners in my book and stashed away to be made again in the future.
Still Life
Last week, about this same time, as I was doing some late-night work at the kitchen table with the sliders open, I heard some rustling in the backyard. Finally, at a few minutes before eleven, I grabbed my flashlight and headed out to discover not one, but two opossums.
"The people who use AI well right now are drawing on judgment they formed before these tools became ambient. They know when to trust an LLM’s output and when to push back because they learned to read, argue, and sustain attention under conditions in which those acts were not so easily outsourced. They bring something the tool cannot supply."
This is so essential. People growing up today with AI who haven't formed their judgment independently of AI will not have this benefit, and that's what frightens me.
That post from Minas Karamanis was entirely written by Claude: https://boxobarks.leaflet.pub/3mj42airv3s2o#fingerprints-of-the-beige-liebox