Interface design can steer behavior.
Defaults, friction, salience, button hierarchy, prompts, subscriptions and recommendation systems are not neutral. The FTC has documented deceptive “dark pattern” designs that manipulate consumer choices.
Summary judgment
Keep agency, exit, portability, interoperability, repair and calm technology as Human Scale values. Stop treating local, open-source, offline, decentralized or quiet systems as automatically more humane.
The real test: legible control, credible exit, proportionate attention, acceptable failure modes, and meaningful human authority over consequential decisions.
What survives strongly
Attention costs should be measured by mechanism.
Interruptions and task switching can impair performance in specific settings, but “screen time” is too broad to function as one psychological exposure.
Portability and interoperability can reduce dependency.
Open standards, exportable data and substitutable interfaces can create credible exit. They still interact with security, privacy, support, intellectual property and usability.
“Human in the loop” is not automatically meaningful oversight.
A reviewer needs actual authority, time, information, competence and the power to disagree or appeal.
What becomes more precise
Quiet technology becomes attention proportional to the task.
An emergency alert should demand attention. A calculator should not. Test task completion, missed information, interruption, errors and time rather than treating silence itself as virtue.
Open source is not identical to user freedom.
Open software can be unusable, insecurely configured, abandoned or dependent on centralized services. Proprietary systems can sometimes provide accessibility, maintenance and security. Measure effective agency.
Repairability matters, with limits.
Repair can extend product life and reduce switching costs, while batteries, high voltage, medical devices, radio certification, calibration and cybersecurity can create legitimate constraints.
Offline capability is not automatically more resilient.
Local fallback can survive network/vendor failure; cloud services can provide replication, remote backup and professional operations. Design around the failure modes that matter.
Automation may preserve or erode skill.
Ask which skills still need practice for judgment, safety, recovery, development or creativity instead of using “deskilling” as an automatic anti-automation argument.
More settings do not automatically create more agency.
Legible defaults, reversibility and credible alternatives may matter more than hundreds of configuration options.
Centralization can create capability and dependency.
Scale can bring expertise, coordination and lower unit costs while concentrating power and correlated failure. Decentralization can distribute control while fragmenting maintenance and accountability.
Recommended chapter changes
Core: Keep “Technology should expand human capability without requiring human captivity.” Define captivity through practical exit, dependency, attention cost, privacy, failure modes and power.
Quiet: Technology should demand attention in proportion to the task and return it when the task is complete.
Open: Prefer standards, portability, interoperability, documented interfaces and repair where they create credible exit without unacceptable safety/privacy/security costs.
AI oversight: A human reviewer counts only when the person has enough information, authority, time and competence to meaningfully disagree.
Automation: Remove drudgery while identifying which human skills still need practice for safety, judgment, development, recovery or dignity.
What Chapter 5 gets right
Technology is a tool, not destiny. Exit, portability, ownership, repair, attention, fallback, judgment and dependency all matter. Convenience can conceal power.
The chapter is weakest when openness, locality, decentralization, quietness or repairability become proxies for human flourishing rather than mechanisms to evaluate.
What would change our minds?
We should weaken preferred principles if closed systems consistently provide much greater safety/accessibility/security without meaningful dependency costs; offline features create more fragility than resilience; calm interfaces cause important information loss; automation reliably increases skill; or centralized services repeatedly outperform distributed alternatives on the actual failure modes that matter.
We should strengthen them where portability, repair, credible exit or calmer interfaces repeatedly produce measurable gains.
Result of Audit 05
Technology should expand human capability while preserving legible control, credible exit, proportionate attention, acceptable failure modes, and real human authority over consequential decisions.The best technology does not have to be small, local, open source, offline or invisible. It has to justify what it asks humans to surrender.