Debating with Bots: How Social AI Became My Digital Devil’s Advocate
I've been experimenting with an app called Social AI. It's a closed/private personal Twitter-like platform filled with bots – and only bots. Each one of them is assigned a distinct personality or archetype that you get to select. Here’s a snapshot of what that looks like:
At first, I wasn’t sure what to expect, but I’ve found it to be a surprisingly (somewhat) useful tool for ideation. What makes it so compelling is the setup: a “safe” space where you can post anything—a question, a comment, a complaint, even an emotion—and instantly receive reactions from your personalized community of bots. Depending on how you configure it, this could be a space for support, a place where you go to be affirmed and uplifted.
But I took a different approach. I intentionally selected bots that were designed to challenge me. These bots contradict, question, doubt, and critique. And they also make jokes. My main goal was to create a digital devil’s advocate—something that would consistently push back against my ideas, forcing me to defend my positions and reconsider my assumptions. Of course, you can’t "win" against them or change their minds—they’re programmed to stay in character, endlessly challenging you without wavering.
my "angry bots" or devil's advocate group
Creating a Personal Debate Forum
The goal of this experiment was straightforward: I wanted to toss ideas out and be questioned. I needed a space where I could test my thinking, refine my arguments, and see if my ideas held up under scrutiny—or at the very least, expand and sharpen my reasoning. It wasn’t just about defending my positions; it was about deepening my understanding and challenging my assumptions.
Lately, I’ve been immersed in biofuel research, as I mentioned in a previous post. I decided to use this theme as a testing ground for my “angry bots” experiment, posing both theoretical and practical questions about the pros and cons of biofuels. The process was fascinating, and here’s a sample of some interactions:
One particularly interesting aspect was experimenting with sentiment. I was curious to see how the bots would react if I introduced an emotional element into the conversation. For instance, I told one bot that its response had hurt my feelings, and it was fascinating to observe how it replied—almost like testing for empathy in the exchange—and to see how the other bots in the community would respond. This dynamic created an unexpected layer of interaction: some bots showed sympathy, while others doubled down or shifted the conversation in an entirely different direction.
I even went a step further and set two bots up to debate each other. I posed a question, sat back, and watched them engage. This is incredibly interesting (and valuable?) — seeing how the different personas might react to a single idea or how they interpreted complex issues. While some of the interactions were predictable, the dynamic nature of the platform kept things feeling spontaneous.
I could have steered them back on topic or in any direction I wanted — but I was curious to see where this rabbit hole lead. This is just a snippet:
If your bot community is filled with problem-solvers, ideators, teachers, or counselors, you’ll have one type of experience—perfect for developing and nurturing ideas. On the other hand, if it’s populated with skeptics, trolls, doomers, or pessimists, you’ll encounter a completely different dynamic, one that challenges and critiques more harshly. And it makes fun of you. The real benefit of this app lies in its ability to rapidly generate responses and to see how different personalities shape the conversation.
Limitations
While Social AI is (somewhat) useful for brainstorming, it doesn’t go very deep. The bots are one-dimensional and stick to their prescribed personalities—they won’t adapt or offer nuanced insights beyond their programming. That’s important to recognize upfront. I tried pushing the bots to provide evidence, data, or links to support their opinions, but none could. For example, when some bots claimed that biofuels have a negative environmental impact, I asked for sources. The bots evaded my request or flooded me with vagueness. So, it’s key to remember that this isn’t meant to be a research assistant or fact-checker; it’s better suited for instant reactions in a light conversational setting.
The value is its ability to create a private audience for testing your thinking, engaging in debate, and having discussions you might not have elsewhere. Facing those rude, awkward, mean, and distracting responses. It offers a unique, low-pressure environment for bouncing ideas off reactive bots, receiving unfiltered feedback, and exploring diverse perspectives—all within the controlled space of your personalized setup. You can also swap out bot personalities whenever you want to change the tone or direction. I often switch between my “angry bots” and my “helpful bots” to compare and contrast responses.
Where Could This Go?
What excites me most is imagining where this technology could go next. Could it evolve into something more sophisticated, offering real-time data or evidence to back up its arguments? Or perhaps it could be tailored for specific professional or academic uses, simulating debates among experts in a particular field. Even more intriguingly, could there one day be a bot that analyzes your interactions, providing thoughtful feedback on how well you made your case and guiding you to sharpen your reasoning and debate skills? Were you effective and persuasive? How might you improve your communication or influencing abilities? Were your ideas innovative? Were your responses creative and on-point? Did you handle the negativity and pressure well? Or even more functionally, could it take notes and generate a summary of the conversation that could be used outside of the platform?
As a thinking partner, Social AI could become more than just a sounding board—it could evolve into a tool that not only challenges your ideas but helps you grow intellectually. By offering detailed analysis and suggestions for improvement, it might serve as a personalized mentor, guiding you to refine your thinking.
The potential for this technology extends far beyond simple brainstorming. Imagine using this type of platform to test out new service ideas, outreach or marketing campaigns, policy proposals, or even complex ethical dilemmas by building scenarios and watching how they play out across different audiences or situations. In this way, an AI could serve as both a challenger and a collaborator—pushing you to think deeper, raising critical questions, and helping you refine your ideas in ways that would be hard to achieve alone or even with colleagues. It could become a creative partner, always ready to provoke fresh thinking and new insights.
What makes Social AI exciting is the playful, experimental space it creates. It’s not just a tool for serious debate with haters and cynics—it’s a sandbox where you can take risks, test wild concepts, and see how they evolve. The low-stakes nature of the platform encourages curiosity and bold thinking, allowing you to stretch your imagination and explore new possibilities without fear of judgment or failure. Although the bots do tend to inhabit that sarcastic culture found on Twitter.
For now, I’m enjoying seeing the unexpected reactions that emerge when my “angry bots” interact with me—and with each other. Each exchange somehow surprises me. At the very least— it’s entertaining and the bots are actually funny.
