On the Monday Lucía opened her job app again after medical leave, she expected an uncomfortable but useful mix of possibilities. During those months she had completed training in cultural mediation, volunteered at her neighbourhood library, and realised she wanted to move away from the phone-based customer service that had filled her recent years. Instead, the screen produced an almost perfect row of telemarketing, call-centre supervision, and telephone sales vacancies. The app seemed certain about her past in a way she no longer was.
It was not an innocent recommendation. Every alert carried the same unspoken message: this is who you are because this is what you did. Lucía could dismiss the cards, but the system kept learning from her old clicks, her previous contract, and the words most common in her résumé. Her new course was reduced to a secondary line. The experience of caring for a relative during leave appeared nowhere. Neither did the harder-to-datafy idea that someone can return to working life with a different pace and different priorities.
A job tool is genuinely helpful when it expands a person’s ability to choose. It stops helping when it turns a recorded work history into a destination. The issue is not that it suggests roles related to previous experience; that clue can be valuable. The issue begins when the assumption becomes invisible, cannot be corrected, and crowds out any new direction before the person has had a chance to explore it. A job seeker needs support translating what they can do, not a machine that endlessly returns an older version of them.
Leave does not erase the possibility of change#
Returning after leave may involve physical recovery, grief, exhaustion, caring responsibilities, or simply a deep reconsideration of what someone wants to sustain every day. There is no single return story. Treating that period as a gap to be filled quickly excludes essential information: someone may have gained a skill, recognised a boundary, or concluded that their former environment is no longer compatible with their health. None of those conclusions needs to become a detailed disclosure to a platform.
Returning after leave may involve physical recovery, grief, exhaustion, caring responsibilities, or simply a deep reconsideration of what someone wants to sustain every day.
The app does not need to know a diagnosis in order to offer respectful options. It only needs to let a person state, at the level of detail they choose, that their goals have changed. “I want roles with less travel,” “I am moving towards cultural work,” or “I need to explore before receiving automatic alerts” are legitimate employment preferences. They are not anomalies that require explanation or defects an algorithm should compensate for.
Bias starts before the recommendation#
Recommendations are built from signals: previous jobs, searches, applications, saved vacancies, and the responses of people with apparently similar profiles. If all those data come from a period before a change, the tool accumulates evidence in favour of the past. Sometimes it adds another layer: it learns which applications receive fast responses and pushes sectors that hire at high volume. In that way it mistakes the likelihood of an immediate interview for the relevance of an opportunity.
For Lucía, telephone-based roles were plentiful and the system had years of data connecting her to them. Cultural-mediation activities appeared less often, required less standardised combinations of skills, and generated fewer clicks. If the platform labels the first group “best match” and hides the second at the end of a list, it is not describing a neutral reality. It is ordering the horizon and making it more costly to look beyond the familiar route.
Give the steering wheel back to the job seeker#
The first change is visible and straightforward: clearly separate history from current goals. A profile screen should ask which experiences a person wants to use as references and which they prefer not to prioritise for now. A former job can remain as evidence of experience without governing every recommendation. Just as someone updates their city or desired working hours, they should be able to update their professional direction.
That editing needs human language. A box labelled “advanced preferences,” hidden behind several menus, is not enough. The app can offer clear statements: “I want to change sector,” “I am interested in related functions, not the same title,” “do not send suggestions based on this job,” or “I am returning to job searching gradually.” Each choice should be reversible and should not demand intimate justification. The person controls the input because they will live with the consequences of the output.
A second change is to explain every suggestion. Instead of a mysterious percentage, a card might say: “We are showing this vacancy because you said you are interested in activity programming, you have experience coordinating schedules, and you saved two museum listings.” Next to that explanation there must be real controls: “this fits,” “show fewer roles like this,” “my goal has changed,” and “show alternative paths.” Explanation is not a decorative sentence; it is a route for correcting the inference.
A personal project does not fit inside a keyword#
Many transitions begin as a still-incomplete sentence: “I want to work with people,” “I am drawn to education,” “I need hours compatible with caring,” or “I want to test something before committing.” A search engine that accepts only exact job titles forces that sentence into a professional label too early. It can be useful to provide a short space for describing a personal project and to treat it as a direction, not as public data for recruiters.
Human guidance is not an emergency button#
Some decisions cannot be resolved by an automated recommendation: how to mention leave without oversharing, which skills to foreground when moving fields, whether an apparently flexible offer has sustainable conditions, or when a search is becoming overwhelming. A good app does not replace those conversations; it makes them easier to reach. It can offer a voluntary referral to careers guidance, unions, public services, or specialist organisations, depending on where the person is and what they need.
That referral should arrive before the system locks someone into a loop. If a user repeatedly rejects the same family of jobs, the answer should not be to persist with nearly identical listings. A quiet invitation could appear: “These suggestions may not reflect what you are looking for. Would you like to review your goals with an adviser or explore transferable skills?” This is not a penalty for declining to click. It is a sign that the product has recognised its own limit.
What to check before trusting a platform#
A job seeker cannot redesign an app alone, but they can check how much room it offers. Look for whether it lets you sort or exclude sectors, edit goals without erasing all experience, save separate searches, and understand why alerts arrive. If a platform persists with an old profile, it can be used as one limited vacancy source while you add sector-specific boards, contacts, training, guidance services, and manual searches. No one is obliged to hand one screen the task of defining their future.
A practical exercise can restore perspective. On a sheet of paper, make three columns: experiences you want to retain, conditions you do not want to repeat, and directions you want to test. These do not need to be finished job titles. “Coordinate,” “explain,” “work near home,” “avoid split shifts,” or “learn in a small team” are already useful signals. Then update the profile with what the system allows and observe what changes for two weeks. Save examples of poorly directed recommendations; they will help you adjust filters or frame a conversation with an adviser.
Lucía did something similar. She kept her customer-facing experience but stopped marking telephone sales as a priority. She added activity coordination, accessibility, and community work; created a wider alert for cultural centres; and booked an appointment with the municipal employment service. The app did not become perfect, but it stopped being the only voice arranging her options. She first found a short library cover role and, from there, gained more precise language for continuing her search.
A responsible platform does not promise to guess anyone’s vocation. It offers revisable maps, explains why it draws a route, and accepts that a person may take a turn. That is the question digital-employment designers should ask: are we using the past to open a conversation, or to close one? The answer appears in concrete controls, not a slogan. For a job seeker, the next step can be small but decisive: recover the ability to say, “This is what I was, but it is not all I may become.”
