Accountability Becomes Inevitable

Imagine applying for an opportunity and receiving a decision based on an account of your life that you have never seen. It includes transactions, messages, movements and associations collected for different purposes over many years. Nothing needs to be newly exposed. The information was already there; what has changed is the ease with which an institution can assemble it and decide what it means.

The ingredients are familiar. A transaction leaves an entry. A service retains an account history. A photograph records a moment that its subjects may later forget. Individually, these traces answer limited questions. A payment can establish that money moved between accounts without explaining why it moved. A location estimate may place a device near a building without establishing who carried it. Combining records can improve an explanation while also allowing one uncertain inference to lend apparent support to another.

Even sources of information that seem remote from ordinary observation can become relevant. Research has demonstrated forms of human sensing from WiFi signals under particular experimental conditions. That is a reason to think carefully about what an environment might disclose. It is not evidence that ordinary networks already provide a universal, accurate account of everyone's movements.

The speculative step is convergence. Suppose institutions acquire the means and permission to connect many such sources, and automated systems make the work inexpensive enough to perform routinely. A message that once sat unread in an archive might be assessed alongside a payment or a later association. An old event could acquire a new significance because the system can now place it beside something else. The past would change in its practical effects without any new action by the person being judged.

This does not require a perfect record to affect behaviour. A person only needs to believe that an action may remain retrievable and consequential. Before sending a message, attending an event or contacting someone, they may consider how the association could look to an unknown future reader. Sometimes that hesitation would prevent harm. Sometimes it would discourage an innocent act whose context might be difficult to establish later.

Convenience could carry much of the adoption. A service that simplifies verification, reduces fraud or makes a decision faster has an immediate attraction. The costs of linking records may appear later, after other institutions have begun expecting access to the same information. A person can consent to one useful arrangement and subsequently find that declining the wider arrangement has become expensive. Dependence can grow through a series of choices that each looked modest when considered alone.

There are substantial reasons to want better records. They can expose abuse, contradict a powerful person's denial and establish that an obligation was ignored. Forgetting is not equally merciful to everyone. Someone whose account was dismissed may need evidence to persist long enough to be heard. A defence of privacy that treats all retrieval as an intrusion would overlook the people for whom the missing record was part of the harm.

The difficulty begins when preservation is mistaken for judgement. Evidence that something happened does not, by itself, settle its significance or the response it warrants. A message may be authentic and still ambiguous. An association may be real and tell us little about agreement. A person's earlier conduct may be relevant without describing their present conduct adequately. A system can store the event faithfully while producing an unfair account of the person.

Nor is the assembled account neutral simply because the inputs are factual. Someone decides which sources count, what associations to infer and how far into the past a decision should reach. Someone decides whether a disputed entry remains visible and whether a later correction follows the original accusation. These choices may be distributed across organisations rather than made by a single intelligence. Their consequences still arrive together for the person whose access is refused.

That is why opportunity, trust and participation could become the main means of enforcement. The consequence need not be a public conviction or a dramatic punishment. An application receives a lower ranking; an account requires additional checks; an invitation never arrives. If the person cannot discover which claim informed the decision, they may have no practical way to correct it. The system's confidence then matters more to their life than their ability to explain themselves.

The freedom at stake is more precise than the freedom to tell any story we like about our past. People remain answerable for harm they caused. They also need ways to contest a false allegation, supply missing context and have evidence of change considered. A history that can only accumulate accusations is not made accurate by becoming longer. Even a correct account of an earlier period requires judgement about its relevance now.

The outcome depends on choices that remain open: what is collected, how long it is retained, who may connect it and what a person can challenge. Technical limits, incompatible systems, institutional resistance and deliberate protections can all alter the path. The existence of scattered capabilities does not establish that they will converge into an unavoidable order. Treating the result as inevitable would make it easier to overlook the decisions producing it.

The possibility that concerns me is therefore a world in which judgement becomes routine before the people subject to it gain a corresponding right to question it. More of the past may become available, and that can make denial harder. Accountability still requires someone to answer for how the record is used. If the person is permanently visible while the decision about them remains beyond examination, we have made surveillance more dependable than responsibility.

Further reading: NIST: AI Risk Management Framework; Geng, Huang and De la Torre: DensePose From WiFi.