The way a problem is named shapes who is held responsible for it. The standard phrase 'vulnerable populations' names the problem in a way that, examined closely, points responsibility in the wrong direction. This article is about why, and about the more accurate framing the Foundation uses instead.
What the standard framing implies
'Vulnerable populations', as the phrase is ordinarily used, names a fixed list of groups understood to be inherently fragile. The grammar of the phrase locates vulnerability inside the people: it is something they are, a property they carry. Three things follow from that framing, and all three are unhelpful.
It implies the condition is permanent, these groups simply are vulnerable, as a stable fact. It implies the condition is total, a vulnerable person is vulnerable, full stop, rather than vulnerable to particular things. And, most consequentially, it implies that if harm occurs, the explanation lies in the fragility of the person rather than in anything about the system that harmed them. If vulnerability is a deficiency the person brought with them, the system that fails them is, in a sense, just encountering a pre-existing problem.
Why that framing is inaccurate
It is not only unhelpful; it is wrong on the facts. Consider the same person in relation to two different AI systems. A person with low digital literacy may be acutely vulnerable to a complex automated benefits-eligibility system and not vulnerable at all to a well-designed voice interface that asks simple questions. Nothing about the person changed between those two cases. The vulnerability appeared in one relationship and not the other.
Consider, too, how people move in and out of vulnerability. A confident, capable adult navigating systems easily on an ordinary day can be placed into a vulnerable relationship with those same systems by a sudden illness, a bereavement, a financial shock, or a crisis that consumes their attention and capacity. Vulnerability arrived without any permanent characteristic of the person being involved.
Both observations point the same way. Vulnerability is not a stable property of the person. It is a property of the relationship between a particular person, in a particular situation, and a particular system.
The relational definition
Stated precisely: a person is algorithmically vulnerable to a system when two conditions hold together. The system's failures would carry serious consequences for them. And they have limited ability to detect, contest, avoid, or recover from those failures. Both conditions describe the relationship, what is at stake for this person in this system, and how much power they have within it, not a trait of the person considered alone.
Why the reframing changes everything
This is not a semantic preference. The relational definition relocates responsibility, and it does so by a simple logic. If vulnerability is a fixed deficiency in the person, then nothing is owed by anyone, the situation is regrettable but no one's doing. If vulnerability is created by the relationship between a person and a system, then it can be changed by changing the system, and the people who design, deploy, and govern that system are precisely the people with the power to change it.
Responsibility, in other words, follows the ability to act. The relational framing makes visible that the system's builders can reduce the vulnerability, through clearer communication, design that does not assume a narrow typical user, genuine recourse, conservative behaviour at high stakes, and therefore makes visible that they have a reason to. The standard framing hides exactly that, by placing the problem somewhere no one can reach it: inside the person.
This is why the Foundation insists on the term algorithmic vulnerability and on the relational definition behind it. It is not a gentler way of saying the same thing. It is a more accurate way, and the accuracy points the obligation at the people who can actually meet it.
