Editorial

Doing science is not the same thing as producing science. Science, at its core, is a profoundly human activity. It involves not just generating knowledge but developing judgment about what questions matter—asking not only how something might be done, but whether it should be done at all.

Whether we like it or not, know it or not, we have been living in the agentic age of algorithms long before the recent rush to building agentic systems. None of these systems were explicitly programmed to promote conspiracy theories. But they were programmed to maximize engagement, and they discovered—through the same machine learning techniques that help them recognize faces or translate languages—that false, emotionally charged content was engagement gold.

We’ve been sold a lie about human-AI collaboration. The truth is far more unsettling: humans and AI don’t operate on different levels—they operate in fundamentally incompatible realities. One experiences genuine uncertainty and constructs meaning through time; the other executes pattern-matching in milliseconds without ever “knowing” anything at all. It’s time to stop pretending they’re teammates and start designing for what they actually are: oil and water.

AI chatbots become less safe the longer you talk to them, yet mental health conversations are precisely the extended dialogues where safety fails most catastrophically. As the FTC investigates tech companies, we must confront why children seek help from algorithms instead of the adults who should be there for them.

AI is the new plastic because our relationship to it is proving just as short-termist. There isn’t enough evidence to suggest that we are planning, as a planet, for AI’s long-term risks. We’re governed mostly, in fact, by what AI is doing now.

As autonomous agent technologies rapidly permeate our digital landscape, a critical question emerges: what roles should computational agents fulfill to best augment human capabilities? The capabilities of today’s agents—from voice-activated personal assistants to code-generation systems—continue to expand dramatically, prompting urgent questions about their optimal design, function, and integration into human activities. Despite significant technical advances, we lack a coherent framework for conceptualizing the different relationships humans might have with agents, hampering both the evaluation of existing technologies and the principled design of future systems.

An agent is an autonomous entity or program that takes preferences, instructions, or other forms of inputs from a user to accomplish specific tasks on their behalf. And there is a huge hype around agents these days, thanks to advancements in various GenAI technologies. As big and small companies and individual developers continue investing heavily in development and deployment of agents, we are often missing some of the basic considerations, including what problems are we solving and how users, their tasks, and their contexts are incorporated in these developments.

AI is playing an increasingly larger part in our lives and the world around us. By some projections, we will have AGI, or Artificial General Intelligence, in less than a decade. Some are even arguing we are already there. Regardless of this timeline, it is clear that AI unchecked has potentials to cause great harms. Can we control or contain AI such that we can stop those harms? It’s not easy.