Following on from my previous blog, I’ve been thinking about relationship issues where ChatGPT is a collaborator in a project rather than simply being a tool. To do this, I’m going to talk first about some common tools – those that serve us without thinking about what they are doing – and the relationship between our expectations of tools, tool use and our performance. I’m then going to look at the kind of relationship issues we might have with ChatGPT as a continually learning team member.
I remember when portable calculators became “a thing” in maths classes at secondary school. To start with, only the “rich” kids had them. Eventually they became something that everybody had to have, in the same way they needed pens and paper. Their main advantage seemed to be that calculations, especially those that involved multiple operations, could be completed more quickly. Improved productivity – got to be good, right…? Right – up to a point…
Faster calculations didn’t necessarily represent more accurate calculations. As a tool, the calculator did not think for itself – its neural pathways were engineered to respond precisely to the inputs it was given, and only those inputs. That’s also where its flaw was – as students we learned to depend on its answers (which must be right, because it’s a machine, after all), but we seemed to lose some of our ability to think critically about why an answer might not be what we expected it to be. In speeding towards the answer, because we wanted to get it quicker than anybody else, we lost focus on the process. We lost the ability to reverse engineer it to check for data entry and operation errors.
Who was the least valuable team member now? The calculator, because it told us what was, even if that wasn’t what we wanted to see? The humans, because our declining ability to critically examine the process meant that we could have created an inevitable wrong solution to a problem? We didn’t think of the calculator as a collaborator in some shared problem-solving process. To us it was simply a tool, in the same category as hammers and screwdrivers – it was something you used to do a job more efficiently than you could do that same job without it.
It’s important to remember in this example that the tool will perform only as well as the operator enables it to. The more experienced the operator, the better results you could expect them to produce using the tool. The value of the experience, of course, comes from the learning connected to previous tasks using the tool. The same tool might help a different operator produce better results than you, but the tool on its own could never outperform you. Our liking, or otherwise, of the tool was based on how well it enabled us to perform. Unless the tool actually broke during use, any relationship dissatisfaction would have been more properly directed at ourselves rather than at our tools.
ChatGPT, as a collaborative project team member, forces us to re-examine our relationship with an increasingly sophisticated tool. Unlike the hammer or the calculator, this tool is one that can outperform its humans. I was tempted to say that it can learn from its mistakes, but that raises the question of whether it actually makes mistakes. The tools already mentioned don’t think about what they’ve done and offer you alternatives where a human decision is required. They don’t allow you to engage with them as ChatGPT does. They simply do what they do, whereas ChatGPT deliberately wants to learn how to do what it does both more efficiently and more effectively.
If ChatGPT is capable of doing directed research more quickly than we can, would that not make it the one member we wanted to have on our project team? Perhaps part of the answer to this relies on the extent to which we come to rely on it to suggest alternative research directions for the problem we’ve set it. If the relationship becomes one of significant or total dependence, where we aren’t exercising appropriately critical thought about what ChatGPT is giving us, the relationship is headed for trouble when we aren’t able to adequately explain or defend something that it’s concocted as an answer for us.
If it’s brave enough to say when it thinks there are several options to effectively answer a question we’ve set it – because it doesn’t have to worry about the emotional reactions of anybody else on the team – would we not be happy about that? Think of the time and emotional energy that could be redirected from intra-team disputes to more productive decision-making and action. On the other hand, not having to observe the human social niceties in interactions with the ChatGPT algorithm could cause us problems if we forget to “switch back” when interacting with our human team members (and project stakeholders outside our team).
Thinking about those human relationships, there are probably always going to be team members that we “like” more or less than others. The justifications are valid for us alone, based on a combination of facts (real or perceived) and suppositions, with maybe some prejudices thrown in for good measure. It’s not uncommon for project team members to socialise outside of the context of the project – from where you get the fun and games that go with who’s in the “in” group and who’s in the “out” group. I’ve seen teams get to the point where some members can’t even be in the same physical space at the same time.
You might argue that’s childish behaviour from adults. Without debating the merits of that argument here, remember that ChatGPT doesn’t have an ego it needs to have stroked. It’s just there to do what it’s been asked to do, to the best of its continually improving ability. Perhaps that changes the heart and soul dynamic of the team, but it’s too early to say if the direction and scope of that change is a good or a bad thing. I think it’s something we are going to have to experience before we can reach that decision, and we should certainly be starting that thinking now rather than leaving it for later.
I’m interested in learning about your thinking on this topic because it’s through collaborative conversations that we can create even more valuable learning and development experiences. Please do respond with a comment and let me know what you think.
In my next blog on Thursday 6 April, I will build on my thoughts here and share more of my thinking about issues associated with ChatGPT, particularly those that might arise when team members are working across different languages.
Rob Nelson Education 