Activity theory: the object of activity and the contradictions that move it

A theory of interaction can take the individual task as its unit and predict a time to completion. Or it can take the collective activity as its unit, and explain why the same tool serves one setting and fails in another. Activity theory, the subject of this fourth of ten units, takes the second route. Its explanatory payload is the object of the activity and the contradictions among the components that pursue it. That payload buys an account that survives the setting's history, at the cost of a frame that describes what happened better than it predicts what will happen.

Why the unit of analysis is an activity, not a task

The tradition begins with mediation. Vygotsky argued that cultural artefacts, tools and signs, mediate action on an object. Internalising social activity, he held, produces distinctively human cognition (Vygotsky, 1978). The isolated mind is therefore not the unit of analysis. Leont'ev added the hierarchy HCI later adopted. Activity is collective and object-oriented, an action is individual and goal-oriented, and an operation is routine and conditioned by the tools at hand. An action is oriented by the collective activity it serves, which is why one keystroke can mean different things in two workplaces (Leont'ev, 1978).

Kuutti (1996) presented that hierarchy to HCI as an alternative to information-processing psychology. On his account, that tradition had attracted criticism for problems in research and design. Nardi's edited volume then assembled thirteen contributions applying the framework to human-computer interaction, moving the work into the field (Nardi, 1996). The unit of analysis is a system with an object, a community, rules, a division of labour and mediating artefacts.

The object is what makes a tool make sense

Bodker (1989) turned the hierarchy into a design position. She asked why a bank teller needs a different interface from a casual user of the same machine. Her answer was that the framework must come from the work situations in which the artefacts are used. A teller works a stream of customer transactions through the machine all day, and a casual user withdraws cash once. A graphic designer and a secretary meet the same text program through different objects. A feature that helps one member of each pair costs the other. A list of operations abstracted from the work will not settle the design. Her conclusion is the unit's sharpest sentence. The user interface cannot be seen independently of the use activity, meaning the users' socially organised practice and its material conditions, including the object of the activity. She states that the article illustrates with examples rather than systematic results.

Engestrom (2000) draws the same line with more precision. He distinguishes short-lived, goal-directed actions from durable, object-oriented activity systems. The prime unit of analysis is the historically evolving collective system, seen in its network relations to other systems. Such systems, he argues, are driven by communal motives that individual participants cannot articulate. A tool can therefore fit every action in a workflow and still cut across the object that gives the workflow its point. Task-level testing will not show that.

Contradictions, and why they are the engine

Engestrom (2000) does not treat trouble as noise. Activity systems are, in his account, in constant movement and internally contradictory. Their contradictions appear as disturbances and mundane innovations, which opens the possibility of an expansive transformation. That transformation runs as a cycle: questioning standard practice, analysing its contradictions, modelling a vision for the zone of proximal development, then implementing the new model. Engestrom (2001) states five principles: the activity system is the unit of analysis, activity is multi-voiced, it has a history, contradictions drive change, and change can take an expansive form. He applies them through four questions about who learns, why, what and how, in a study of expansive learning in a Finnish hospital. About 60 physicians, nurses, other staff and managers came from the Children's Hospital in Helsinki and from primary care health centres. They met in 10 three-hour sessions that ran into February 1998, seated on opposite sides of the room by organisation. They watched videotapes of children whose care had drifted between providers. They named one contradiction between an object that crossed organisations, the child's care trajectory, and rules of cost-efficiency enforced inside each one. A second lay between that object and instruments built for care within a single institution.

Fu (2025) reports a twenty-month investigation of the Mathematics Stack Exchange community, using participant observation, an analysis of 1,942 governance discussion threads and interviews with 28 contributors. The study identifies four contradictions: between tools and objectives, between rules and objectives, between division of labour and objectives, and within the objectives themselves. Three resolution mechanisms appear: user-created tools that work around technical limits, flexible governance through tagging, and moderation roles balancing quality control against newcomer support. The paper is the unit's best evidence that the concept works on a live setting, cautioned by one thing: the contradictions were named after the analysis.

Why a tool that transforms one setting fails in another

Engestrom (2001) reports that the third generation of the theory takes two interacting activity systems as its minimal unit, to address inter-organisational learning. Engestrom and Sannino (2010) review the resulting work across six strands, from object transformation and cycles of learning actions to boundary crossing and formative interventions. They add a section on the critiques. Across a network of interconnected systems, they note, objects are only partially shared and often contested.

Engestrom and Sannino (2020) then chart four generations of the tradition through the widening of its unit of analysis. They run from mediated action, to the collective activity system, to multiple interconnected systems, and most recently to heterogeneous work coalitions aimed at wicked societal problems. The four generations co-exist, they observe, and the theory must now develop the fourth. The tool travels, the object does not, and the contradictions that surface are the mismatch appearing as conflict, workaround or quiet abandonment.

The price: a frame that describes more than it predicts

Roth and Lee (2007) present cultural-historical activity theory as a theory for praxis, which changes how research proceeds rather than issuing forecasts. Sannino and Engestrom (2018) set out the Helsinki school's two guiding principles: double stimulation and ascending from the abstract to the concrete. They conclude that these make formative intervention the powerful way to conduct activity-theoretical research. Engestrom and Sannino (2010) state the criterion directly. The ultimate test of a learning theory is how far it helps practitioners generate learning that grasps pressing issues.

The price is that the framework's decisive evidence sits inside an intervention run by the people whose practice is being changed. A description that fits the setting it was written in cannot be checked before the setting is entered, which the predictive tradition refuses.

Contested question: is the object of activity an explanation or a placeholder?

The literature disagrees about its central term. The first position holds that the object organises the system, and that contradictions involving it explain where change will come from. Engestrom (2000, 2001) state it, and Fu (2025) is the current case. The second comes from inside the tradition. Kaptelinin (2005) argues that the concept's usefulness is undermined by unsolved problems in its meaning and contexts of use. He identifies three sources of uncertainty. They are the translation of Leont'ev's Russian into English, rival interpretations of the concept in the Leont'ev and Engestrom traditions, and Leont'ev's own definition of the object as the activity's true motive. Kaptelinin would separate that last definition from the motive. If the framework's central term is the term under dispute, two analysts can attribute different objects to one setting and each remain internally consistent.

The third position is methodological. Engestrom and Sannino (2010) hold practitioner usefulness to be the ultimate test. The information-processing tradition that Kuutti (1996) documents as the criticised mainstream requires a theory to state performance in advance. What would settle it is a criterion for individuating the object of an activity, fixed before data collection. A prediction would also be needed, from the object-and-contradiction analysis, that a task-level analysis of the same setting does not make. A study in which two analysts applied the criterion independently to one interaction and reported whether their objects agreed would test the concept as Kaptelinin asks. Until then the object is an explanatory commitment rather than a measured variable.

The evidence is uneven in kind. The system side offers two case studies, a hospital intervention with about 60 participants and a twenty-month study of one online community. In both, the contradictions were named after the analysis. Kaptelinin's (2005) objection is conceptual, since it identifies confusions in the term without case data behind it. While the dispute is open, a designer should write down the object of the activity before the fieldwork starts. The brief should say in advance what finding would show the object had been misnamed. The strongest objection is that the explanatory payload arrives after the event. Fu (2025) named the four contradictions after the analysis, so an account that explains the change that happened cannot say in advance which change is coming.

Where the argument is heading

Two named lines are live. The first extends the framework to settings where a model computes. Adeel and Apperley (2026) propose a multidimensional adaptation they call Pyramidal Activity Theory for interaction with digital twins. The adaptation incorporates users, models as tools, live data that is unavailable offline, interfaces and outcomes. They demonstrate it through two conceptual case studies, one mapping interactions in advanced manufacturing and one informing a prototype digital-twin interface. The authors state that empirical evaluation remains future work, so the extension is a proposal rather than a result.

The second line is the tradition's own reorganisation. Engestrom and Sannino (2020) argue that the fourth generation, the heterogeneous coalition, must now be developed. Sannino and Engestrom (2018) close on the major challenges facing the theory this century. The direction is outward, towards many actors and contested objects, and inward, towards the object concept in systems whose components compute. Kaptelinin's objection travels with both, and the most useful next paper may be the one that answers it with a criterion rather than another case.

The sources this unit rests on

• Vygotsky, L. S. (1978). Sets out the cultural-historical account of higher psychological functions, in which mediation by tools and signs and the internalisation of social activity produce distinctively human cognition. It is the origin of the unit's claim that artefact and action must be analysed together.
• Leont'ev, A. N. (1978). In Activity, Consciousness, and Personality (Prentice-Hall), he sets out the hierarchy of activity, action and operation, and the claim that an individual action is oriented by the collective activity it serves. The unit uses it for the three levels that keep an activity apart from a task.
• Kuutti, K. (1996). Presents activity theory to HCI as a candidate alternative to information-processing psychology, at a point when that mainstream framework had come under criticism for problems in research and practical design. It carries the tradition's entry into the field and the three-level vocabulary in HCI terms.
• Nardi, B. A. (ed.) (1996). Context and Consciousness (MIT Press) is the edited volume that assembled thirteen contributions applying activity theory to human-computer interaction, with a deliberate move out of the laboratory into the field. It is the signpost for the framework's establishment in interface research.
• Bodker, S. (1989). Argues that the user interface cannot be seen independently of the use activity, meaning the users' socially organised practice and its material conditions, including the object of the activity. It supplies the unit's clearest design position, and states that it illustrates with examples rather than systematic empirical results.
• Engestrom, Y. (2000). Distinguishes short-lived goal-directed actions from durable object-oriented activity systems, takes the system as the prime unit of analysis, and argues that systems are internally contradictory, with contradictions appearing as disturbances and mundane innovations. It gives the unit both its definition of the object and its expansive cycle.
• Engestrom, Y. (2001). States five principles of activity theory, adds four questions about learning, and makes two interacting activity systems the minimal unit of the third generation, applying the scheme in a Finnish hospital study. It is the unit's compact statement of the framework.
• Engestrom, Y. and Sannino, A. (2010). Reviews expansive-learning studies across six strands, from object transformation to formative interventions, and includes a section on the critiques. It states the unit's price directly: the ultimate test of a learning theory is how far it helps practitioners generate learning that grasps pressing issues.
• Engestrom, Y. and Sannino, A. (2020). Charts four generations of the Finnish school through the widening unit of analysis, from mediated action to heterogeneous coalitions for wicked problems, and finds that the generations co-exist and the fourth needs development. It carries the unit's account of why the object travels less well than the tool.
• Fu, H. (2025). Reports a twenty-month study of Mathematics Stack Exchange using participant observation, 1,942 governance threads and 28 contributor interviews, and identifies four contradictions and three resolution mechanisms. It is the unit's strongest empirical test, and its own caution, since the contradictions were named after the analysis.
• Kaptelinin, V. (2005). Argues that the usefulness of the object of activity is undermined by unsolved problems in its meaning and contexts of use, identifies translation and rival interpretations as sources of uncertainty, and calls for separating the object of activity from the motive. It carries the unit's principal internal objection.
• Roth, W. M. and Lee, Y. J. (2007). Presents cultural-historical activity theory as a theory for praxis that transforms how research proceeds, drawing on Vygotsky and the later generations. It carries the unit's claim that the framework's aim is practice rather than prediction.
• Sannino, A. and Engestrom, Y. (2018). Sets out the Helsinki school's central ideas, including the guiding principles of double stimulation and ascending from the abstract to the concrete, and argues for formative intervention as the way to conduct activity-theoretical research. It carries the unit's account of the framework's method and its current challenges.
• Adeel, A. and Apperley, M. (2026). Proposes a multidimensional adaptation of activity theory for digital twins, incorporating users, models, live data, interfaces and outcomes, demonstrated through two conceptual case studies. The authors state that empirical evaluation and usability assessment remain future work, so it marks direction rather than result.