Everyday Robots in Everyday Life

Introduction

Over the past few weeks, I've had multiple conversations with robotics entrepreneurs about evolving physical AI systems from specialized skills to more general capabilities. This prompted me to go down a rabbit hole of advanced manufacturing robotics video demonstrations. These “collaborative” robots (cobots) are undeniably technically impressive. They’re also socially oblivious. While the robots performed their programmed tasks effectively in isolation from their human colleagues, I witnessed broader communication and coordination issues on the shop floor that could result in workflow disruption, hazards, and potential safety shutdowns. What struck me wasn't just a usability issue. Cobots generally have sensors that allow them to detect and react to human presence and movements. But these sensors don’t prevent social performance breakdown – robots have no idea how to play their part in changing social dynamics.

This insight crystallized a question I've been grappling with: How do we design physical AI systems that function not just as technical tools but as social actors in our complex human environments?

As AI-powered physical systems increasingly populate our factories, hospitals, roads, and homes, we find ourselves at a critical juncture. These embodied intelligences must navigate both functional requirements and social contexts – a challenge that calls for integrating insights from seemingly disparate disciplines.

Two theoretical frameworks offer particularly powerful, complementary perspectives. The sociologist Erving Goffman's approach to social interaction examines how humans manage impressions and perform roles across different contexts. User Experience guru Don Norman's user-centered design principles focus on creating intuitive objects that align with human cognitive patterns. When brought together, these viewpoints create a robust framework for developing physical AI that succeeds both technically and socially.

But why does this matter? Because as AI systems physically enter our world, their social integration – not just their technical capability – will determine whether they enhance or disrupt our lives. Are we prepared to design not just for function, but for social performance?

Theoretical Foundations: A Conversation Between Sociology and Design

Goffman's Social Stage

In 1959, Goffman published "The Presentation of Self in Everyday Life," introducing a perspective that would transform our understanding of social interaction. His framework uses theatrical metaphors to analyze how we present ourselves in daily life.

Goffman's core insight – that social interaction is a performance where individuals manage impressions according to context – provides an informative lens for understanding human-AI interaction. Several key concepts are particularly relevant:

Impression management describes how we actively control others' perceptions through appearance, manner, and setting. Consider how a doctor’s white coat and formal language establish trust and authority, or how a teacher's tone shifts between encouragement and discipline depending on the situation. Physical AI systems, through their design and behavior, inevitably create impressions that shape how we respond to them.

The distinction between front stage and backstage behavior highlights how our performances differ in public versus private settings. We present polished versions of ourselves in professional contexts but relax these performances in trusted environments. Should robots maintain consistent performance regardless of context, or should they adapt to different "stages"?

Face-work refers to strategies we employ to maintain social dignity and help others preserve theirs. When someone makes an error in conversation, we often pretend not to notice or offer face-saving responses. Do AI systems need similar capabilities to smooth over social disruptions?

Norman's Design Wisdom

Don Norman's influential work "The Design of Everyday Things" (1988) revolutionized how we think about the objects we use daily. His principles emphasize creating intuitive relationships between form, understanding, and function:

Affordances and signifiers are properties that indicate how an object can be used. A door handle's shape suggests it should be pulled; a plate on a door suggests pushing. For physical AI, these concepts highlight the importance of designing robots whose form suggests their capabilities and limitations. Does a healthcare robot's appearance accurately signal what it can and cannot do?

Feedback mechanisms provide information about the results of our actions. When you press a button and see a light change, you know your action had an effect. In AI systems, feedback becomes crucial for establishing trust and enabling smooth interaction. How do we design robots that unambiguously communicate their understanding and intentions?

Conceptual models are the mental representations users develop about how systems work. Norman emphasizes that good design creates accurate conceptual models that help predict system behavior. For robots, this means creating systems whose operation aligns with our intuitive understanding of social actors.

Where These Worlds Intersect

At first glance, Goffman's sociological theory and Norman's design principles might seem to occupy separate domains – one focused on social performance, the other on object usability. But for physical AI, these perspectives converge.

Both frameworks emphasize contextual behavior – how actions must be tailored to specific environments. For Goffman, this means modifying social performances based on the audience; for Norman, it means designing objects whose use fits their context.

They offer complementary perspectives on failure. Goffman examines how we recover from social disruptions, while Norman explores how good design prevents technical failures. Physical AI must manage both dimensions simultaneously.

Both recognize the tension between visible and invisible structures. Goffman examines unspoken social rules, while Norman discusses how design principles can be communicated or embedded. AI systems must navigate both explicit and implicit aspects of interaction.

This integration raises profound questions: How do we design robots that are both technically usable and socially appropriate? Can we create machines that understand not just what to do, but how to perform their functions in ways that respect social norms?

Applications: Designing for Function and Performance

Creating Socially Legible Robots

Imagine walking into a hospital and encountering a robot in the corridor. How would you know what it does? Whether it's safe to approach? If it might move suddenly? These questions highlight the need for physical AI to be both functionally clear and socially legible.

The integration of Goffman and Norman suggests robots should be designed so their physical form (e.g., size, shape, movement patterns) communicates not just functional capabilities but social behavior. This isn't merely about aesthetics; it's about creating machines whose appearance establishes appropriate expectations.

A hospital delivery robot designed with Norman's principles would have intuitive controls, reliable path-finding, and secure cargo handling. But applying Goffman's insights would additionally consider how the robot's height, speed, and communication style establish appropriate social presence. Should it move differently around patients? Should it announce itself before entering rooms? Should it acknowledge hospital hierarchies?

The Bauhaus movement taught us that "form follows function," but for social robots, we might say "form communicates performance." The physical design of a robot implicitly signals its social role and behavioral expectations – a responsibility designers must take seriously.

Contextual Intelligence: Reading the Room

Most people behave in a library versus a sports stadium. Our communication style tends to shift between a formal meeting and a casual dinner. This adaptability to social context is so natural for us that we rarely think about it, yet it poses an arguably existential challenge for physical AI.

Goffman's front-stage/backstage distinction offers a valuable framework for programming contextual intelligence in robots. Consider the cobot working alongside its human colleagues on the factory floor. It might operate differently during:

• Production time (front stage): Moving efficiently but predictably, using clear signaling behaviors, maintaining professional distance

• Maintenance periods (backstage): Allowing closer human proximity, simplified movement patterns, and more direct feedback

Norman's principles of mapping and feedback can ensure these contextual shifts are transparent to users. Visual indicators might signal when the robot is switching operational modes, while consistent response patterns help workers understand the robot's current state and intentions.

The question isn't simply whether a robot can perform its technical function, but whether it can adapt that performance to different social contexts in ways that humans can intuitively understand. This represents a shift from programming robots to perform tasks to programming them to perform roles.

Failure Recovery: When Things Go Wrong

Those awkward social moments we’ve all experienced reveal something important: social success isn't about avoiding all errors (impossible even for us humans) but about recovering gracefully when they occur.

For physical AI, failure can come in two forms, technical and social. Both require careful design attention. Norman's approach to error handling emphasizes preventing mistakes, making them reversible, and providing clear recovery paths. Goffman's concept of face-work suggests social disruptions should be managed in ways that preserve dignity.

Together, these perspectives suggest designing robots that can:

• Recognize when they've made a technical or social error

• Acknowledge the error appropriately (neither ignoring nor overemphasizing it)

• Offer clear paths to resolution that minimize embarrassment

• Learn from the interaction to prevent similar failures

Consider an autonomous vehicle that needs to make an unexpected stop. The technical aspect involves safe braking and system diagnostics. But the social dimension is equally important: How does the vehicle communicate the situation to passengers? How does it signal its intentions to other drivers? A well-designed system handles both the functional emergency and the social disruption it creates.

This dual approach to failure reveals a broader truth: in physical AI, technical and social design aren't separate considerations but deeply intertwined aspects of the same challenge.

Case Studies: Theory Meets Practice

Healthcare Robots: Trust in Vulnerable Contexts

Healthcare environments present particularly complex challenges for physical AI, combining high-stakes technical requirements with sensitive social dynamics.

In healthcare settings, robots must navigate contexts where patients are vulnerable, privacy concerns are heightened, and established medical hierarchies exist. The integration of Norman and Goffman suggests healthcare robots should feature:

• Clear technical affordances that communicate medical capabilities without overpromising

• Social performances that establish appropriate professional distance while conveying care

• Context sensitivity that distinguishes between routine care, emergencies, and private moments

• Failure recovery mechanisms that prioritize patient safety while maintaining trust

A robot designed to assist with physical therapy, for instance, would need interfaces that clearly communicate how patients should interact with it physically. Simultaneously, its social performance (communication style, proxemic behavior, responsiveness) should establish an appropriate therapeutic relationship that motivates without intimidating.

The stakes in healthcare are particularly high because trust is essential to effective care. A robot that functions perfectly but violates social norms or fails to acknowledge patient dignity could undermine the very outcomes it's designed to achieve.

Manufacturing AI: Collaboration Without Threat

In industrial settings, collaborative robots work alongside human operators in shared spaces. These environments present unique challenges where safety, efficiency, and worker comfort must be balanced.

For manufacturing AI applications, our integrated framework suggests:

• Design affordances that clearly communicate operational capabilities and physical limitations

• Social performances that establish appropriate collaborative relationships without intimidation

• Context-sensitive behaviors that distinguish between routine operations and exceptional situations

• Interaction patterns that respect worker expertise while providing assistance

The factory-floor cobot exemplifies these challenges. Applying Norman's principles, its physical design should make its range of motion and functional capabilities immediately apparent through form, lighting systems, and interface elements. Its controls should be intuitive and provide clear feedback about operational status.

From Goffman's perspective, the robot's movements and signaling behaviors should establish appropriate "social distance"—moving confidently during routine operations but adapting its speed and proximity when humans enter its workspace. Its performance should acknowledge the expertise hierarchy of the environment, perhaps deferring to experienced operators while providing more guidance to novices or visitors.

The cobot must also manage impression formation carefully. It should appear competent without threatening worker status or job security, presenting itself as a tool for augmentation rather than replacement. This might be achieved through collaborative behaviors that visibly depend on human input for certain decisions, reinforcing the complementary nature of the human-robot team.

Historical parallels exist in how early industrial automation was designed. The most successful deployments weren't those that simply replaced human labor but those that thoughtfully integrated into existing workflows and social structures. Are we learning these lessons as we deploy increasingly autonomous systems in workplaces?

Automotive AI: Social Navigation on Public Roads

Autonomous vehicles represent perhaps the most publicly visible form of physical AI, operating in complex social environments governed by both formal rules and informal social norms. When I first rode in a self-driving car prototype, I was less concerned with its technical functioning than with how it would navigate the unwritten rules of the road – the subtle negotiations between drivers at four-way stops, the implicit communication through slight speed changes or positioning, and pedestrians!

A Goffman-Norman framework suggests that automotive AI should incorporate:

• Intuitive interfaces that communicate vehicle capabilities and current operational mode

• Social signaling that makes vehicle intentions legible to other road users

• Context-sensitive behavior that adapts to different driving environments and cultural norms

• Graceful transitions between autonomous and human control

The self-driving vehicle must manage multiple types of interaction simultaneously. For the passenger, Norman's principles suggest designing interfaces that create accurate mental models of the car's capabilities and current state, clearly distinguishing between different levels of autonomy and providing appropriate feedback about decision-making processes.

For other road users, Goffman's perspective becomes crucial. The vehicle must perform its intentions in socially legible ways—communicating when it will yield, merge, or turn through movements that human drivers can interpret. It must also recognize and respond to the social signals of others, reading the subtle cues that indicate another driver's intentions beyond formal signaling.

The autonomous vehicle also faces complex impression management challenges. Its driving style inevitably creates impressions such as appearing cautious or confident, deferential or assertive, that affect how others respond to it. Designers must consider what "personality" their vehicle projects through its movement patterns and how this affects both passenger comfort and social integration with other road users.

This raises profound questions: Is driving primarily a technical activity or a social one? How do we design vehicles that navigate roads and the social spaces around them? And how might the introduction of autonomous vehicles transform the social norms of driving itself?

Ethical Considerations: Designing with Responsibility

Authentic vs. Strategic Interaction

When Siri responds to my requests with an enthusiastic tone, is it being authentic or merely performing happiness? Although we intuitively understand it’s performative, the application of Goffman's theory to AI design raises important questions about authenticity and manipulation.

If robots are programmed to perform certain social roles (appearing to be caring, attentive, or deferential), is this fundamentally deceptive? Norman's emphasis on transparent design suggests robots should clearly communicate their actual capabilities and limitations. Yet Goffman's work reminds us that all social interactions involve some degree of performance and impression management.

The challenge is finding an ethical balance between creating socially fluent AI and avoiding manipulation. One approach is designing robots that are honest about their nature while still being socially adept—systems that don't pretend to have emotions or understanding they lack, but still honor social conventions that make interaction comfortable for humans.

Consider a social robot that uses gentle, caring language. It needn't claim to "feel" concern, but its behavior can still respect human dignity and emotional needs. This distinction between simulating emotion and demonstrating care through behavior offers a path forward.

As designers, we must ask: What responsibilities do we have when creating machines that perform social roles? How transparent should we be about the performance aspects of AI behavior? And how do we balance social fluency with honest representation of what AI systems actually are?

Power Dynamics and Agency

Physical design and social performance can reinforce or challenge power dynamics. The height of a robot, its voice characteristics, and its programmed behaviors can all signal social status and authority, impacting human agency in the interaction.

Norman's principles emphasize putting control in users' hands, while Goffman's analysis highlights how social interactions often involve subtle power negotiations. For ethical AI design, this suggests creating systems that:

• Provide clear user control over robot functions

• Avoid reinforcing harmful stereotypes through design or performance choices

• Consider how robot behavior might affect different user groups differently

• Allow users to customize the social aspects of interaction (remember TARS from Interstellar?)

As physical AI enters workplaces, homes, and public spaces, we must ask: How do design choices affect the balance of power between humans and machines? Are we creating systems that enhance human agency or subtly diminish it? And how can we ensure physical AI serves the needs of diverse users rather than reinforcing existing inequalities?

Future Directions: The Road Ahead

Measuring Social Performance Success

How do we know if a robot is socially successful? While technical functionality can be measured straightforwardly, evaluating social performance is more complex. A robot might complete its tasks perfectly while creating awkward, uncomfortable, or inappropriate social interactions.

Future research needs to develop metrics that consider not just whether a robot completes its tasks, but how well it integrates socially. This might include:

• Observational studies of human-robot interaction over extended periods

• Measures of social acceptance and comfort with robot presence

• Analysis of how robot performance affects human-to-human social dynamics

• Longitudinal studies of attitude changes toward robots over time

This challenge has historical parallels in how we've assessed other technologies. Early computer interfaces were evaluated primarily on technical metrics like processing speed, but the field evolved to include comprehensive usability and user experience measures. Social robotics may require a similar evolution, combining "Can it do the task?" with "Does it do the task in a way that integrates well with human social environments?"

Co-Design: Bringing Together Technical and Social Expertise

The integration of technical and social perspectives suggests the need for collaborative design approaches that bring together engineers, designers, social scientists, and end users. The siloed nature of current development processes, where technical capabilities are designed first, with social considerations added later, often results in systems that function but don't integrate well socially.

What might a truly integrated design process look like? It could involve:

• Ethnographic research to understand social contexts before design begins

• Inclusion of sociologists and anthropologists on design teams

• Iterative testing that evaluates both functional usability and social integration

• User feedback mechanisms focused specifically on social dimensions of interaction

This approach would mirror the evolution of user experience design in digital products, where initial technical-focused development gradually incorporated a deeper understanding of human psychology, needs, and behavior (thanks, in large part, to Norman’s contribution).

Adaptive Social Learning

Perhaps the most promising direction is developing physical AI systems that can refine their social performance over time based on interaction experience. Systems that learn technical tasks and social appropriateness.

This would involve:

• Algorithms that recognize social feedback (both explicit and implicit)

• Learning systems that modify behavior based on social success

• Methods for sharing social learning across multiple robots

• Frameworks for balancing adaptation with predictability

Such systems would embody the ultimate integration of Goffman and Norman—robots that learn both how to function effectively and how to perform appropriately in social contexts.

Conclusion: Designing for a Shared Future

The evolution of physical AI from specialized tools to generalized capabilities requires more than just enhanced algorithms and sensors. It demands a fundamental rethinking of what constitutes "intelligence" in embodied systems. By bringing together Goffman's sociological framework and Norman's design principles, we create a path toward robots that succeed both as functional tools and as social actors. Systems that understand not just what to do but how to be in human spaces.

This integrated perspective recognizes that robots are simultaneously technical artifacts constrained by design and social entities engaged in performance. Their effectiveness will increasingly depend on their ability to navigate both dimensions fluidly and appropriately.

For engineers and designers, this requires a paradigm shift. Social performance can’t be treated as an afterthought or superficial layer added once technical capabilities are established. Instead, we need interdisciplinary development approaches that recognize the inseparable nature of function and social integration from the earliest design stages.

The implications extend beyond individual products to how we conceptualize AI advancement itself. Perhaps the path to more generalized AI capabilities isn't just through more sophisticated algorithms but through systems that understand the social dimensions of intelligence, such as the ability to read contexts, adapt performances, and participate appropriately in the complex choreography of human interaction.

We have a choice about how physical and embodied AI evolves. Will we create technically impressive but socially awkward systems that disrupt our environments, or will we design technologies that enhance human experience by integrating thoughtfully into our social fabric? The answer will shape not just the future of technology but the texture of everyday life in a world increasingly shared with intelligent machines.