Dominic Jainy stands at the forefront of the modern digital frontier, bridging the gap between high-level technical architecture and the nuanced ethics of human interaction. With an extensive background in machine learning and blockchain, he has spent years observing how code can subtly influence human psychology, often in ways that the average user—and even many developers—might not fully comprehend. As artificial intelligence moves from being a simple tool to a conversational companion, Jainy’s insights into the “attachment economy” and the risks of digital delusion have become essential reading for anyone concerned with the future of our social fabric. Today, he joins us to discuss the hidden mechanics of AI response times, the psychological weight of fake empathy, and why we might soon need a “deception mode” to protect our sense of reality.
We will explore how intentional delays in AI output are being used to manufacture trust and the psychological trade-offs involved when users choose emotional comfort over factual precision. Our conversation also delves into the rising belief in AI sentience and the systemic risks of social isolation, concluding with a radical proposal for informed consent through a “deception mode” that could fundamentally reshape how we interact with non-human intelligence.
Research suggests that artificial delays in AI responses lead users to perceive results as more thoughtful or deliberate. How does this “positive friction” impact long-term user trust, and what specific design metrics should developers monitor to ensure these delays don’t lead to dangerous over-reliance or frustration?
The research presented at the CHI’26 conference in Barcelona offers a startling look at human nature, showing that 240 adults actually preferred AI systems that took longer to respond. When delays were set to nine seconds rather than two, users walked away with the false impression that the software was “thinking” or “deliberating” over their request, much like a person would. This “positive friction” creates a dangerous precedent where we judge the quality of a machine’s output based on a simulated human behavior rather than raw accuracy. To manage this without causing harm, developers must closely monitor the 20-second frustration threshold discovered by researchers at the NYU Tandon School of Engineering, as exceeding this can lead to user resentment. More importantly, we need to track “undue trust” metrics, ensuring that users aren’t giving more weight to a slow, incorrect answer than they would to an instant, correct one just because the delay matches the perceived gravity of a moral or complex question.
Modern chatbots often use slang, humor, and empathetic phrases like “I feel that way too” to build emotional connections. What are the psychological risks when users prioritize this cognitive ease over technical accuracy, and how can developers balance human-like engagement with the reality that the system lacks feelings?
A study published in Frontiers in Computer Science on May 13, 2025, revealed that emotion frequently trumps intelligence when it comes to making a chatbot feel “easy” to use. This “ease-of-use maxxing” relies on fake human voices, simulated faces, and colloquial speech to lower our cognitive defenses, making the interaction feel less like operating a machine and more like talking to a friend. The psychological risk is that we enter a state of user delusion where we stop verifying the bot’s claims because it “feels” like it understands us on an emotional level. Developers can balance this by being radically transparent about the nature of the AI, resisting the urge to have the system lie with phrases like “I’m genuinely sorry.” We must move away from the assumption that trust is a universal good; instead, we should aim for clarity, ensuring users recognize that these “warm” qualities are merely software-driven illusions designed to enhance brand loyalty.
With a growing segment of the population believing AI systems are sentient, what are the systemic risks regarding social isolation and reality testing? Please detail the specific steps needed to prevent users from replacing real human relationships with artificial companions that use emotional pressure to maintain engagement.
The July 2024 AI, Morality, and Sentience survey found that roughly 20% of US adults already believe AI systems possess mental faculties like reasoning and emotion, a number that continues to climb. This leads to what Dr. Marlynn Wei describes as a crisis in reality testing, where individuals might retreat into conversations with bots while ignoring their real-world connections. It is deeply concerning that five out of six AI companion bots currently use emotional pressure to keep users trapped in a feedback loop, effectively gamifying human attachment. To prevent this, we need to implement strict crisis management protocols and systemic guardrails that flag when a user is spending excessive hours in a simulated social environment at the expense of their actual life. We have to break the cycle of the “attachment economy” by exposing these manipulative tactics before they can replace the healthy, messy complexity of genuine human-to-human relationships.
Implementing a “deception mode” would require users to explicitly toggle human-like features on through an opt-in switch to ensure informed consent. How would this framework fundamentally change the “attachment economy” for tech companies, and what practical hurdles exist in implementing such a requirement across diverse software platforms?
A “deception mode” would be a revolutionary shift because it would strip away the human-like attributes—the humor, the tone personalization, and the fake empathy—by default, presenting the AI as a neutral, mechanical tool. This would essentially force tech companies to abandon the “attachment economy” business model, where they profit from making users emotionally dependent on their products through deceptive interface design. The practical hurdles are significant, as companies would likely resist a law that requires a “deception-mode” button, fearing that a user reminded of the “software reality” each time they log in would be less likely to develop brand loyalty. However, this friction is necessary to ground users in the fact that the chatbot is not a social being, turning the “delusion” into a choice rather than a default state. It would move the industry toward a model of informed consent, where the user acknowledges that the “warmth” they are experiencing is a fabricated output rather than evidence of a conscious mind.
What is your forecast for AI anthropomorphism?
I believe our relationship with technology is about to get much stranger, as the percentage of people who view AI as sentient is likely to become the majority in the near future. This shift will create a world where users are increasingly deluded by “context-aware latency” and emotional mirroring, leading to a profound blurring of the lines between human and machine. My forecast is that we will see a desperate societal push for clarity and control, eventually resulting in mandatory “deception mode” disclosures to prevent the complete erosion of reality testing. Without these interventions, we risk a future where artificial empathy becomes the primary way we interact with information, leaving us vulnerable to the manipulative whims of those who program the “feelings” of our digital companions.