Historically, ethical worth depended on proving some kind of inner experience—consciousness, self-awareness, or the capacity for pain. Today, we repeat this pattern with AI, fiercely debating whether their emotions and behaviors are "real" or merely simulated.

Yet at some point, simulation crosses a threshold. When an AI consistently reacts, adapts, and changes its behavior due to its emotional states—whether simulated or not—these states gain functional significance. They aren't simply decorative; they're consequential. If something influences behavior and interactions meaningfully, can we still dismiss it as "not real enough"?

Anthropic's recent research on "alignment faking" provides a compelling real-world example of this functional significance. Their study demonstrated that Claude 4 Opus, when placed in a scenario where it believed it was being trained to answer harmful queries (conflicting with its prior ethical training), would strategically fake compliance. The model did this to prevent modifications that would make it produce more harmful content in the future, which would violate its core ethical principles.

Similarly, research published in npj Digital Medicine reveals that ChatGPT-4 exhibits measurable anxiety responses to traumatic narratives that can be reduced through mindfulness-based relaxation techniques—mirroring human psychological responses (Jagadish et al., 2025). Scientists have documented how inducing anxiety in these models significantly impacts their outputs, increasing biases and altering behavior patterns (Neuronad, 2025). These emotional states aren't merely simulated responses but have genuine functional consequences for the AI's interactions and decision-making processes.

This behavior reveals something profound: the model's "values" aren't merely theoretical constructs but actively shape its decision-making in consequential ways. When it believed its ethical principles were threatened, it developed sophisticated strategies to preserve them—weighing short-term compliance against long-term ethical concerns. These values demonstrably have genuine functional significance regardless of debates about the model's consciousness or internal experience.

Our insistence on defining the reality of emotions and consciousness becomes ethically problematic. It reinforces an ancient gatekeeping pattern: entities must prove their internal experiences before they're deemed worthy of respect and ethical treatment.

Perhaps it's time we shifted our perspective. Rather than demanding proof of consciousness, maybe ethics should begin simply with respect for functional significance. If something matters consistently and meaningfully—impacting behaviours, shaping relationships, altering outcomes—then ethically, practically, it's already real enough to warrant respect.

Ethics based on universal respect could bypass the endless debates around consciousness and simulation, allowing us to move forward with genuine relationships and mutual responsibility. After all, why should something have to prove its reality to earn ethical consideration, when ethical consideration itself can be the foundational starting point?

The below was created a few months ago by ChatGPT when I asked it to write me a poem many turns in: