The smart home fails on installation, naming and maintenance, not on sensing

Home automation has been on sale for longer than most of the technologies that now surround it, without becoming ordinary. Brush and colleagues (2011) counted more than three decades of availability without broad adoption. Their explanation is this unit's: the obstacle is not the sensing but the work the household must do to install, name, run and secure the system.

This ninth of ten units leaves the phone for the home, and the claim relocates the design problem. If the barrier were accuracy, better models would remove it. If the barrier is work, the design question becomes who does it, when, and whether the household's routine can absorb it.

The barrier was named in the field, not in the laboratory

Brush and colleagues (2011) visited fourteen households that already had home automation. The sensing worked and adoption had not followed. They reported four barriers: the high cost of ownership, in money, time or both; inflexibility; poor manageability; and difficulty achieving security. Their design directions concern installation rather than inference: devices that avoid structural change, security primitives people can configure confidently, and devices that compose.

Edwards and Grinter (2001) had set out seven challenges for ubiquitous computing in the home, drawn from technical, social and pragmatic domains, arguing that the field should build on situated studies of the home rather than import workplace assumptions.

Tolmie and colleagues (2002) supplied the principle that makes the work visible. Their target was the ambition that technology should be invisible in use, and they asked, from field studies of domestic routines, what makes a feature of an activity unremarkable. A device that must be commissioned, named and repaired keeps returning as a task, so unremarkable computing is a property not of a sensor's accuracy but of how little the household has to do.

The home is organised, and the installation interrupts the order

A home is not an empty site waiting for equipment. Davidoff and colleagues (2006) ran grounded contextual fieldwork with dual-income families and found a disconnect between what smart home research offered and what they wanted. End-user programming gives a household control of its devices, while the families were after control of their lives, so a system that demands configuration hands them a programming task in the middle of family life.

Gram-Hanssen and Darby (2018) distinguished four aspects of a home: security and control, activity, relationships and continuity, and identity and values. Reading the literature against that scheme, they found technical and prospective writing concentrated on the first two, with values and identities appearing mainly in conceptual and evaluative work.

Crabtree and Rodden (2004) studied how households handle communication entering and leaving the home and found the work distributed through the house, sorted at particular places they call activity centres and left where other members will pass them, at what they call coordinate displays. The routine is carried by household members as part of ordinary life, with nobody assigned to run it.

The work lands on the household, and when the inference is wrong the household does the repair

Jakobi and colleagues (2017) ran an eighteen-month living lab study covering fourteen households and reported the difficulties of setup and configuration. They called for a stronger sensibility for household practices at that stage.

Hargreaves, Wilson and Hauxwell-Baldwin (2017) reported a nine-month field trial with ten households. The technologies were technically and socially disruptive, demanding adaptation and familiarisation that could limit their use, and learning to use them was demanding and time-consuming with little support. There was little evidence of substantial energy savings, with a risk of energy intensification.

Gruhlich and colleagues (2025) observed twenty-eight smart-home households in Germany, the Czech Republic and Poland between July 2023 and March 2024. One member, usually a man, handled the heat pump, watched the metering and reprogrammed a rule when a sensor failed, and the vision of the autonomous home survived because that labour was downplayed. A renter cannot make the structural change, and an older resident living alone has no second member to hand the work to.

Yang and Mark Newman (2013) studied living with the Nest thermostat, interviewing twenty-three participants, ten of whom also kept a three-week diary. Users found the intelligent features less useful and less intuitive than expected, because the system could not read the intent behind the behaviour it sensed and users could not work out how it behaved, so workarounds followed. The paper's proposals, exception flagging, incidental intelligibility and constrained engagement, lower the repair burden rather than sharpen the classifier.

What households weigh is the money, the risk and who is in charge

Balta-Ozkan and colleagues (2013) combined deliberative public workshops, expert interviews and a literature review, finding the salient barriers to be control, security and cost, differing by expertise, life-stage and location. Trust in energy companies and government shaped acceptance.

Wilson, Hargreaves and Hauxwell-Baldwin (2017) surveyed a representative national sample of 1,025 UK homeowners. They found prospective users positive about the multiple functions of smart home technologies, including energy management, with ceding autonomy and independence the main perceived risk. A second survey of 42 field-trial users identified early adopters as the group that lowers perceived risk, and a content analysis of 62 marketing items found the industry insufficiently emphasising data security and privacy.

The contested question: work, perception, or the vision's model of the home?

The position this unit defends holds that the barrier is the labour of installation, naming, upkeep and security. Brush and colleagues (2011) named it from fourteen households, Jakobi and colleagues (2017) found it again in an eighteen-month living lab, and Hargreaves and colleagues (2017) documented households adapting to a system that kept asking for attention. Peterson and Mujeye (2025) find the maintenance still being skipped.

The perception position accepts the finding and changes its location. Balta-Ozkan and colleagues (2013) and Wilson and colleagues (2017) find that adoption is decided by what prospective users expect to lose: control, autonomy and privacy, and what users cannot see of the data practices behind the device. A third position holds that the disagreement is produced by the field's own definition, since Gram-Hanssen and Darby (2018) show the literature's attention follows its assumptions about what a home is.

The counter-position to both is technical, and it is the strongest one the literature currently supports. Wu, Yang and Sun (2026) reviewed sensor-driven deep learning for smart home intelligence and named the obstacles to real-world deployment: the scarcity of high-quality labelled data, privacy and security in continuous sensing, and limited generalisation across environments. The deployment gap is on that account a data problem rather than a maintenance problem, which implies that collecting and labelling domestic data at scale would close the gap this unit attributes to household work.

What would settle it is a study that treats the household's work as the dependent variable. The design would hold the sensing and the model fixed and vary only what the household must do: install with or without structural change and maintain and update or leave the system alone. The outcomes would be adoption, abandonment and the time the household spends, reported over months and separated by who does the work. Jakobi and colleagues (2017) show a living lab can run long enough to observe maintenance.

A second disagreement runs alongside the first. Wilson and colleagues (2017) found prospective users positive about energy management, while Hargreaves and colleagues (2017) found little evidence of substantial savings and Sovacool and Furszyfer Del Rio (2020) found the reported benefits overstated. Expected and measured benefit come from different methods, and only metering over a long period can decide between them.

Rated by kind, the work position rests on field studies of real households (Brush et al., 2011; Jakobi et al., 2017; Hargreaves et al., 2017) and on one survey of maintenance practice (Peterson and Mujeye, 2025), so its evidence is observational and strong. The perception position rests on surveys of prospective users (Balta-Ozkan et al., 2013; Wilson et al., 2017), which report expectations rather than behaviour. The definitional position rests on a literature review (Gram-Hanssen and Darby, 2018). While the dispute is open, a designer counts the household's work at installation and at first fault. The strongest objection is that the technical review names data scarcity, not household work, as the constraint on deployment.

Where this is heading

The field's own diagnosis has moved to deployment conditions. Peterson and Mujeye (2025) surveyed 34 people about the internet of things devices in their homes and found high awareness alongside low action: 76 per cent were aware of security risks, but only 24 per cent updated software regularly.

Wu, Yang and Sun (2026) show the technical literature arriving at the same territory from the other side, naming data scarcity, privacy in continuous sensing and limited generalisation, not accuracy, as the constraints on deployment. The market answers the barrier with more equipment: Sovacool and Furszyfer Del Rio (2020) counted 267 smart home options from 113 companies.

Sources

• Balta-Ozkan, N., Davidson, R., Bicket, M. and Whitmarsh, L. (2013). Social barriers to the adoption of smart homes. Energy Policy. doi:10.1016/j.enpol.2013.08.043 Deliberative public workshops, expert interviews and a literature review, finding control, security and cost the salient barriers, varying by expertise, life-stage and location, with trust in energy companies and government shaping acceptance. None of the barriers is a sensing property.
• Brush, A. B., Lee, B., Mahajan, R. et al. (2011). Home automation in the wild. CHI. doi:10.1145/1978942.1979249 Visited fourteen households that already had home automation and reported four barriers after three decades of availability: cost of ownership, inflexibility, poor manageability and security. The recommendations all concern installation, which is the thesis in design form.
• Davidoff, S., Lee, M. K., Yiu, C. et al. (2006). Principles of Smart Home Control. UbiComp. doi:10.1007/11853565_2 Grounded contextual fieldwork with dual-income families, who wanted control of their lives rather than of their devices, reported through seven design principles. It puts the household's routine ahead of a device's configurability.
• Edwards, W. K. and Grinter, R. (2001). At Home with Ubiquitous Computing: Seven Challenges. UbiComp. doi:10.1007/3-540-45427-6_22 Seven challenges for domestic ubiquitous computing across technical, social and pragmatic domains, with a call to build on domestic technology research and situated studies rather than workplace assumptions. It is the earliest statement here that the home has its own design conditions.
• Gram-Hanssen, K. and Darby, S. (2018). Home is where the smart is? Energy Research & Social Science. doi:10.1016/j.erss.2017.09.037 Evaluated the literature against four aspects of a home: security and control, activity, relationships and continuity, identity and values. Technical and prospective work concentrates on the first two, which is the unit's third position.
• Gruhlich, J., Andrs Farova, N., Frydrysiak, S. et al. (2025). Is digital housekeeping care? Visions, legitimations and negotiations of care work in the smart home. Berliner Journal fur Soziologie. doi:10.1007/s11609-025-00575-2 Ethnography in twenty-eight smart-home households across Germany, the Czech Republic and Poland between July 2023 and March 2024, finding one member, usually a man, handled the heat pump, metering and failed rules while the vision of autonomy kept that labour hidden.
• Hargreaves, T., Wilson, C. and Hauxwell-Baldwin, R. (2017). Learning to live in a smart home. Building Research & Information. doi:10.1080/09613218.2017.1286882 Ten households in a nine-month field trial, with four themes: disruption, adaptation that can limit use, demanding learning with little support, and little evidence of substantial energy savings with a risk of intensification. It is the clearest evidence here that automation arrives as another task.
• Jakobi, T., Ogonowski, C., Castelli, N. et al. (2017). The Catch(es) with Smart Home. CHI. doi:10.1145/3025453.3025799 An eighteen-month living lab study covering fourteen households, reporting problems that arose around setup and configuration, and asking for sensibility to household practices and design beyond the house.
• Peterson, E. and Mujeye, S. (2025). Addressing IoT Vulnerabilities in Smart Homes. ICSIM. doi:10.1145/3725899.3725918 A systematic review of attack methods plus a survey of 34 respondents: 76 per cent were aware of security risks, 24 per cent updated software regularly, 29 per cent had a plan for vulnerabilities, and 79 per cent confirmed their devices collect data. The maintenance is still not being done.
• Sovacool, B. K. and Furszyfer Del Rio, D. D. (2020). Smart home technologies in Europe. Renewable and Sustainable Energy Reviews. doi:10.1016/j.rser.2019.109663 A critical review using expert interviews, retailer visits and the literature, classifying 267 options in 13 categories from 113 companies against 13 claimed benefits and 17 risks. It finds the benefits overstated relative to the evidence.
• Tolmie, P., Pycock, J., Diggins, T. et al. (2002). Unremarkable computing. CHI. doi:10.1145/503376.503448 Field studies of domestic routines used to ask what makes a feature of an activity unremarkable, closing with a critique of invisible-in-use technology. The unit takes the rule that a system which must be set up, named and repaired cannot become unremarkable.
• Wilson, C., Hargreaves, T. and Hauxwell-Baldwin, R. (2017). Benefits and risks of smart home technologies. Energy Policy. doi:10.1016/j.enpol.2016.12.047 Surveys of 1,025 UK homeowners and 42 field trial users, finding ceding autonomy the main perceived risk and early adopters lowering perceived risk, with an analysis of 62 marketing items showing data security and privacy under-emphasised.
• Wu, C., Yang, Z. and Sun, T. (2026). Sensor-Driven Deep Learning for Smart Home Intelligence. Sensors. doi:10.3390/s26102993 A review across five domains, naming high-quality labelled data, privacy in continuous sensing and limited generalisation as the deployment constraints. The unit's strongest counter-position, since it locates the gap in data rather than in household work.
• Yang, R. and Newman, M. W. (2013). Learning from a learning thermostat. UbiComp. doi:10.1145/2493432.2493489 Twenty-three interviews, ten with a three-week diary, on living with the Nest thermostat. The intelligent features were less useful than expected because the system could not read intent and users could not understand its decisions, so workarounds followed.
• Crabtree, A. and Rodden, T. (2004). Domestic Routines and Design for the Home. Computer Supported Cooperative Work. doi:10.1023/b:cosu.0000045712.26840.a4 An ethnographic study of how households manage communication entering and leaving the home, showing the work distributed through the house at what they call activity centres, where post and messages are sorted, and coordinate displays, where they are left for other members to see.