Bo Bennett On Building Scroops With Psychology

HOST: It's good to have you back, Bo. Last time we mentioned you've been using ScienceBasedLearning for eight months, and you said something that stuck with me: that you built it because every other solution was designed to keep people subscribing, not to actually teach them. Today's new project feels like that same instinct applied to social skills—Scroops.com. And I want to start with the mechanism, because you have a PhD in social psychology and this product is explicitly built around that research. So let's not skip past it: which specific body of social psychology work is load-bearing here? GUEST: Well, mostly we're talking about social skills. We're talking about human interaction, people being able to carry on a conversation and ask the right questions to keep the conversation going and to keep the other party interested. And as you can imagine, this is very important in not only social situations, but business situations as well. So, essentially, it's the science of communication that we're dealing with. HOST: You mentioned reciprocity in that grading rubric—is that based on the idea of balanced contributions to the conversation, or something deeper like emotional mirroring? What's the actual mechanism there that you're trying to train? GUEST: It's both actually. We're talking about uh keeping the conversation two-sided. You hear very often about one-sided conversations. Um whenever you're in a group, there's usually one person that will dominate the conversation. And most of the time they don't realize it. It's just kind of natural styles. But with this system, it kind of trains you in a way where it'll tell you like, okay, you spoke uh 40% of the time or 60% of the time, to give you some idea so people are aware of how much speaking they're doing versus how much listening they're doing. And you you mention emotional mirroring, that's an important part of it as well. If you just start responding to somebody, and uh you you don't really address the thing they said before. That's pretty poor emotional skills, emotional mirroring, you could call it. It It's important in conversation to respond to what the person said before you go on with something that you said, or at least some way acknowledge what they've said. Uh very often people have their own dialogues in their mind, and they know what they want to say next, which means that they're completely not listening to what the other person is saying. This is very bad communication, and this system helps train you in order to pay attention to what the person is saying before you start developing what you want to say. HOST: You're bringing up theory of mind, which is a fascinating phenomenon in terms of human behavior. Now, users describe the person they're about to talk to before the conversation starts. Was that a deliberate social psychology move, or did it come out of the technical requirements of making the AI play a consistent character? GUEST: Again, you could argue that it's both. Uh it it is important to create a character. Um but it's the the character isn't only for technical reasons, it's for practical reasons. For example, if you're um if you want practice, for example, uh going out on a date, then that's very different than practicing for a job interview. So, that's where this this character role comes in. You get to describe exactly the character, the the person that you want to speak with. So, you could practice the conversation as it is most likely going to go. Um whereas practicing just some kind of generic conversation is okay, but we do that every day. It's not good enough. Uh if you want really good practice for let's say uh trying to speak to a mechanic about getting the best price for repairs, then that's the situation that you create. HOST: So, you're explicitly framing this as deliberate practice, much like in sports or music. There is research on anxiety and performance suggesting some level of arousal actually improves social situations. If Scroops eliminates anxiety by making it "safe," do you worry you might accidentally train a performance that's too calm for the real thing? GUEST: No, that's not going to happen because there's always that barrier between a human being and a computer. So, in this case, we're training with a computer. We're talking to a computer, which you have that safety net of not physically being present with the other person. Which really does take away some of the anxiety, but it also takes away some of the um, some of the excitement, uh, the arousal that you mentioned. So, when an actual person is there, you have that extra element. So, it's not, it's not like if you train on the computer 100 times, then the 101st time with a human is going to be exactly the same. There's always going to be that new element, uh, which is going to put you in a lot better position, but it's also going to allow you to have that arousal in order to have peak performance. HOST: So, the safety net of practicing with AI is the key differentiator. Now, you put salary negotiation, breakups, and customer complaints behind the Coach tier. That's a deliberate ranking of consequence. As a social psychologist, what makes those scenarios categorically different from, say, a first date or a networking conversation, and is the AI actually equipped to handle that level of emotional register? GUEST: Well, mostly it's a it's a business decision rather than a science decision to put those uh in a different category. And that's because once we get into um some some more of the the the business aspects of it. Then um then we go to a different tier. Uh but you mention also the um emotional register and no, there's no difference in terms of the technology. Whether you're using a user is using the free version or the um the the paid version, the highest paid version. The AI is trained the same and it's has the same emotional register uh throughout. HOST: So, it's a business decision primarily. Scroops grades users on ten axes after every conversation. If you ran yourself through the Scroops rubric right now—not as the builder, but as the person who has to go have a hard conversation this week—which of those ten axes would come back lowest, and what would the coaching report tell you to work on? GUEST: Well, it depends on what kind of mood I'm in, which is funny cuz sometimes I'm I'm just not very talkative and I don't feel like listening and that's accurately reflected in the rubric. And it's funny how how it could really pick that up. Um, cuz I I recognize it when I'm practicing and I'm playing around with it and I'm I'm doing the programming and the training, I could recognize my lack of social skills, let's say, uh, when I don't have it turned on. Uh, I'm the type of person who who can turn it on at at will, but most of the time as a a programmer and somebody who generally likes solitude, um, I have it turned off. And that that's easily reflected in the rubric. HOST: That's a fascinating insight, recognizing that you can toggle those social skills. It really highlights the potential for deliberate training. Bo, this has been genuinely different from any conversation we've had on this show—which is fitting, given the topic. Let's get you back on soon—there's a thread here I want to pull on once Scroops has some real user data: whether the people who practice the hardest conversations actually go have them, or whether rehearsal becomes its own kind of avoidance. That's a question for a future episode. Before we go—for listeners who want to follow up on what we covered, where can they find you and the work you're doing? GUEST: Well, the site we're talking about is scroops.com, that's scroops.com. And otherwise, you could find all of our websites at rgboy.com. HOST: Bo, thanks for being here and for that nuanced perspective on how your design choices intersect with human behavior. And thank you for spending part of your day with us. Let's get you back on again soon—there's plenty more we can cover. Until next time—that's a wrap.