Image generators like Midjourney aren't just automating artistic labor; they're automating the deeper human capacity to build social connections, according to new critical research. A comprehensive theoretical framework published by researchers analyzing generative models argues that tools for creating images, text, and video are fundamentally changing how we relate to one another by replacing what scholars call "social doing," the effort we exert to build relationships with others.

The research introduces a concept called "Synthetic Sociality," describing a social reality partly fabricated by privately owned generative models that operate without democratic oversight. This goes beyond concerns about job displacement or artistic authenticity; it's about who controls the infrastructure of human connection itself.

What Exactly Is "Social Doing" and Why Does It Matter?

"Social doing" is defined as the effort we invest to build a connection with another person. When you spend time crafting a thoughtful message to a friend, creating a personalized gift, or commissioning custom artwork, you're engaging in social doing. The problem emerges when generative models like Midjourney replace that effort entirely.

Researchers distinguish between two types of automation happening with image generators and similar tools. Substitutive automation means a tool completely replaces the social effort a human would have made. Mediative automation, which better describes how image generators work, partially replaces or supplements social effort while also giving users new capacities they couldn't have accessed alone.

When you use Midjourney to generate a birthday gift image instead of commissioning an artist or creating something yourself, you're engaging in mediative automation. The tool replaces the social doing you might have invested while simultaneously giving you access to artistic capabilities you may not possess. This sounds efficient, but the research suggests there's a deeper cost.

How Did We Get Here? The History of Commodifying Connection

This didn't happen overnight. The research traces how sociality itself became a commodity in the digital economy. Social media platforms like Facebook, Instagram, and TikTok didn't invent the commodification of social connection, but they industrialized it. Every interaction, every message, every moment of connection became data that could be collected, analyzed, and sold.

That accumulation of social data became the raw material for training generative models. Midjourney, DALL-E, and similar tools were trained on billions of images and descriptions that humans created while engaging in social doing on digital platforms. The models learned to replicate not just visual patterns, but the social context embedded in those images.

In essence, generative models automate social doing by using "dead labor," the accumulated social effort of countless people who shared their creative work online, often without understanding how it would be used. This creates what researchers call "commodity fetishism," where the connection between the tool and the human effort behind it becomes invisible.

Steps to Understanding the Broader Implications of Synthetic Sociality

• Recognize the Substitution: Image generators replace not just artistic labor but the social effort involved in commissioning, collaborating with, or learning from human creators. When you use Midjourney instead of working with an artist, you're automating away a relationship.
• Understand the Data Source: The training data for these models comes from billions of social interactions on digital platforms, representing accumulated human effort that was commodified without explicit consent or compensation to the original creators.
• Consider Ownership and Control: Private companies own and govern these generative models with minimal transparency or democratic input. This means a small number of corporations control the infrastructure through which billions of people now engage in social connection.
• Examine the Perception Shift: As we increasingly interact with AI-generated content and use these tools to mediate our relationships, our perception of what's real, authentic, and valuable begins to shift in ways we may not fully recognize.

What Are the Real Consequences of Automating Social Connection?

The research identifies several downstream effects that emerge from automating social doing. First, there's the problem of invisibility. When you use Midjourney to create an image, the tool obscures the connection to the living social doing embedded in its training data. You don't see the artists whose work trained the model, the social context of the images, or the cultural history behind visual styles the model learned to replicate.

This invisibility creates what researchers call "commodity fetishism," where the social and cultural bonds that made the training data valuable become hidden. You see a beautiful generated image but not the thousands of human relationships, artistic traditions, and social interactions that made it possible.

Second, there's the question of who benefits. When Midjourney users generate images without commissioning human artists, the economic value that might have flowed to creators instead flows to the company that owns the model. More subtly, the cultural knowledge and artistic traditions embedded in the training data are extracted and privatized by corporations.

Third, and perhaps most concerning, is the alteration of our social fabric itself. As more people use generative models to mediate their relationships, create gifts, communicate with loved ones, and express themselves, the nature of those relationships changes. We're not just automating labor; we're automating the very capacities that make us social beings.

Can Automation of Social Connection Ever Be Beneficial?

The research doesn't argue that all automation of social doing is inherently harmful. There may be scenarios where it's beneficial, such as helping people with disabilities engage in social connection they might otherwise struggle to achieve, or providing access to creative tools for people without resources to hire professionals.

However, the current deployment of generative models like Midjourney happens without this kind of intentional design for social benefit. Instead, these tools are optimized for profit and user engagement, not for strengthening human relationships or democratizing access to creative capacity.

The research suggests that future generative models could be designed differently. Rather than being privately owned and governed, they could operate with greater transparency and democratic control. Rather than obscuring the social doing in their training data, they could make visible the artists, creators, and communities whose work made them possible. Rather than replacing human connection, they could be designed to enhance it.

As Midjourney and similar tools become increasingly integrated into how we work, create, and relate to one another, these questions become more urgent. We're not just deciding whether to use a new technology; we're deciding what kind of social fabric we want to live in, and whether the automation of human connection should remain in private hands.