The Metaphors We Code AIs By

Can a machine be a friend, a therapist, or a life partner? For a growing number of people, these questions are no longer hypothetical. From virtual therapists to digital soulmates, AI companions are becoming fixtures in our social landscape. This trend, more than just a curiosity, highlights the importance of the metaphors we use to understand artificial intelligence. As this transformative technology rapidly advances, a key challenge lies in grasping its capabilities and limitations. To overcome this challenge, we often rely on metaphors – cognitive frameworks that compare AI to familiar concepts like brains, tools, or partners – to conceptualize and communicate its potential use cases.

Metaphors are powerful; they color our perceptions and expectations. While they shape how we think, feel, and act, can also oversimplify and mislead our understanding of AI. For example, when we talk about AI as a therapist, we might imagine a kind of wisdom, empathy, and emotional reciprocity that it currently does not truly possess. It processes data and patterns but without real understanding or emotional depth.

Consider the most widely used metaphor of AI as a "brain." We see this metaphor reflected in popular culture, from the sentient robots of science fiction to the anthropomorphized virtual assistants in our homes and phones. While the brain metaphor has been instrumental in advancing AI research, it also obscures the fundamental differences between human and machine cognition. By focusing on replicating the biological structures of the human brain, we risk neglecting alternative computational approaches that might better capture the unique strengths of artificial intelligence. By anthropomorphizing AI systems, we risk attributing them to human-like autonomy, consciousness, and moral reasoning–– which could blur the lines of responsibility and accountability when things go wrong.

Similarly, framing AI solely as a tool risks underestimating its potential for autonomy and unintended consequences. The tool metaphor emphasizes human control and agency in shaping AI's development, but it may not adequately capture the ways in which AI systems can evolve and make decisions in ways that are difficult to predict or control.

Another guiding metaphor for AI development is that of a collaborator or teammate. This metaphor highlights the technology's potential to work alongside humans in pursuit of shared goals. However, this metaphor can be equally misleading if taken too literally. AI systems, while capable of impressive collaboration, lack the true agency and moral responsibility we associate with human partners.

Image source: DALL-E

One promising alternative is to view AI as an infrastructure—much like the electricity that powers our homes or the water systems that run beneath our streets. This perspective helps us recognize its fundamental role in our lives and the need for robustness, reliability, and fairness. It encourages us to ensure that AI systems, like any other public utility, are designed with the public good in mind, a shift in thinking that could have profound implications for AI research and policy.

Our choice of metaphors is often influenced by our biases and agendas. Those who stand to benefit from AI's rapid development may favor metaphors that emphasize its potential benefits, while those wary of its disruptive power may gravitate towards metaphors that highlight its risks and uncertainties. In essence, our metaphors can become self-fulfilling prophecies, possibly steering AI development in no clear direction.

Ultimately, the metaphors we choose to conceptualize AI will shape its technical development and social and ethical implications. By being intentional and reflective about our metaphorical thinking, we can work to create AI systems that are more transparent, accountable, and aligned with human values. This will require discussions and collaboration across disciplines, from computer science and engineering to philosophy, psychology, and the social sciences.

Standing at the golden hour of the AI-powered future, let us choose our metaphors wisely. Let us seek metaphors that inspire us to create AI systems that are intelligent, wise, transparent, collaborative, rational, and beneficial to humanity as we know it and what it could be. And let us remember that, just as we shape AI through our metaphors and mental models, AI will also shape us profoundly and unpredictably. It is up to us to ensure that this transformative technology serves the greater good, starting with the language and concepts we use to understand it.