interviews · Frontiers in sociology · la publicación, 4 ago 2026 · gratis
"Ya no me piden que piense. Me piden que esté de acuerdo"
Un estudio entrevistó a 32 profesionales inmigrantes de 50 a 55 años en cuatro sectores de Europa occidental. Describe un patrón; no mide despidos ni sueldos.
Versión breve · la versión detallada sigue, unos 8 min
Pregunte al chat de weeklyAI
Pregúnteme sobre la IA en su vida: le señalo las notas que weeklyAI ha publicado sobre el tema, esta semana y antes.
Las conversaciones se guardan mientras exista weeklyAI, para mejorar la publicación. Se responde en el idioma en que usted escribe.
- El estudio, de un vistazo
- Quiénes
- Profesionales inmigrantes de 50 a 55 años en salud, finanzas, mercadotecnia y tecnología
- Cuántos
- 32 profesionales, ocho por sector
- Dónde
- Varios países de Europa occidental
- Cuándo
- Publicado en agosto de 2026
- Tipo de estudio
- interviews
- Quién lo hizo
- Instituto Jozef Stefan de Liubliana, Eslovenia
- El límite que importa
- Solo 32 personas en cuatro sectores de oficina de Europa occidental; no mide sueldos ni despidos.
El patrón que estos relatos dibujan no es el del trabajador que pierde el puesto, sino el del que conserva el sueldo y pierde la voz.
Un profesional de la salud lo dijo en una entrevista: "A veces siento que ya ni me piden que piense. Me piden que esté de acuerdo". La inteligencia artificial lee primero las imágenes y él confirma lo que encontró.
Los investigadores entrevistaron a 32 profesionales nacidos en el extranjero, de 50 a 55 años, en salud, finanzas, mercadotecnia y tecnología, en varios países de Europa occidental. Fueron ocho por sector y cada entrevista duró alrededor de 55 minutos. Todos llevaban al menos diez años viviendo y trabajando en el país.
El patrón que describieron: conservaron el empleo y el puesto, pero su criterio fue reemplazado por aprobar lo que la máquina ya produjo. Pasaron de autores a revisores. Una enfermera contó que su intuición clínica ahora tenía que discutir con la luz verde de la máquina, y que esa luz verde era la verdad por defecto.
Algunos dijeron que su propia experiencia entrenó a los sistemas que después los dejaron a un lado. Un analista de mercados emergentes contó que pasó 16 años construyendo conocimiento de los mercados financieros del este de Europa, y que esa ventaja se disolvió cuando la IA produjo cobertura parecida a escala. "La máquina se llevó la buena mitad de ser extranjero y me dejó la mala", dijo.
Son 32 entrevistas en cuatro sectores que dependen del conocimiento: salud, finanzas, mercadotecnia y tecnología. El estudio no midió sueldos ni despidos y no puede decir cuántos trabajadores viven esto ni qué pasará con el suyo. Los propios autores llaman a este patrón un marco por probar, no una predicción.
Varios contaron que la capacitación en IA se dirigió a los más jóvenes. Un radiólogo escuchó, de pasillo, que los programas se priorizaban para personal más joven. Otros dijeron que expresar dudas los marcó como resistentes al cambio. Hubo quienes quedaron protegidos: por el tipo de puesto, por la red de contactos o por una empresa internacional.
En su propio trabajo, fíjese cuándo su día empieza a consistir en aprobar resultados que usted no produjo. Los autores piden a las empresas que ofrezcan capacitación en IA también a los trabajadores mayores y extranjeros, y que abran canales seguros para plantear dudas.
Qué significa para usted
Lo que se describe viene de 32 entrevistas en cuatro sectores de Europa occidental, así que no sirve para anticipar lo que pasará en su empleo ni para decir cuántos viven algo parecido. En su trabajo puede observar si su día empieza a consistir en aprobar resultados que usted no produjo. Pregunte también a qué trabajadores se está capacitando en las nuevas herramientas y quién queda fuera.
Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614
Quién pagó: El trabajo fue financiado por la Agencia Eslovena de Investigación con la subvención P2-0037; el artículo no declara relaciones comerciales o financieras que constituyan conflicto de interés y no dice nada sobre el papel de los financiadores en el diseño ni sobre equipos o programas cedidos.
No tome esto como consejo médico profesional.
Versión detalladaLos pasajes copiados del artículo, las ilustraciones y cada fuente con cuánto leímos de ella · 1566 palabras · unos 8 minLeerla →Cerrar
Le queda el oficio, pero ya no la voz: lo que 32 profesionistas inmigrantes contaron sobre la IA en su trabajo
Un estudio cualitativo en cuatro sectores de Europa occidental describe el paso de experto a revisor de la máquina, y nombra el patrón que lo vuelve más pesado a los 50.
Imagine la sala de lectura de un hospital. La inteligencia artificial ya miró las placas y escribió lo que encontró. A usted le toca confirmar. Un radiólogo sénior lo dijo así: "a veces siento que ya ni me piden pensar. Me piden estar de acuerdo"1. Ese es el hallazgo central de un estudio publicado en agosto en Frontiers in Sociology, firmado por Martin Mihajlov y Tanja Pavleska, del Instituto Jozef Stefan de Liubliana, y pagado por la agencia de investigación de Eslovenia2.
Los autores entrevistaron a 32 profesionales nacidos fuera del país donde trabajan, de entre 50 y 55 años, ocho por sector: salud, finanzas, mercadotecnia y tecnología. Todos llevaban al menos diez años radicando y trabajando en algún país de Europa occidental, y venían del sur y del este del continente. Las entrevistas duraron en promedio 55.4 minutos y se analizaron con un método que busca temas repetidos en lo que la gente cuenta. Los participantes también llenaron dos cuestionarios, uno sobre actitud hacia la IA y otro sobre miedo a perder el empleo, que se usaron solo para dar contexto a las entrevistas2.
Salieron diez temas. Los autores llaman al principal Doble Desplazamiento: primero, la exclusión de siempre, la de los años de migrante que nunca termina de pertenecer; después, el golpe nuevo, el que convierte al experto en validador pasivo de lo que produce la máquina3. Trece participantes quedaron en el grupo de desplazamiento fuerte, doce en el medio y siete en el mínimo. Es una clasificación de 32 personas, no un censo: no dice cuántos trabajadores viven esto en ningún país.
El segundo hallazgo tiene nombre propio: la prima de extranjería. Lo que la migración le dio a alguien como ventaja profesional —conocer mercados regionales, moverse entre culturas, dominar dos maneras de decir las cosas— resulta ser justo lo que el sistema aprende a imitar. Una analista de mercados del este de Europa pasó 16 años construyendo su ventaja y la vio disolverse cuando la plataforma cubrió lo mismo a escala: "la máquina se llevó la buena mitad de ser extranjera y me dejó la mala"45.
Hubo un mecanismo que se repitió. La experiencia de estos profesionistas no solo fue rebasada: sirvió de material de entrenamiento. "Mi experiencia fue tratada como datos de entrenamiento. Como una razón para mejorar la máquina para que no me necesite la próxima vez"6. Y hubo un límite en el acceso a la actualización: a un radiólogo le dijeron, casi en el pasillo, que los programas de capacitación en IA estaban "priorizados para el personal joven"7.
Hacia el final de la carrera, el miedo se junta. "El miedo a la pensión y el miedo a la IA son el mismo miedo, la verdad. Los dos apuntan al mismo precipicio"8. Para quien migró hay una capa extra: cotizaciones partidas entre dos países, dos mitades incompletas que no se pueden unir9. Y el silencio: "el sistema está construido para que los más afectados no puedan hablar con seguridad, y luego su silencio se usa en su contra"10.
El estudio tiene límites que importan. Solo son 32 personas, en cuatro sectores de oficina y en Europa occidental; no dice nada medido sobre oficios, trabajo manual ni América Latina. Los propios autores describen el Doble Desplazamiento como un marco analítico que espera prueba empírica, no como un concepto ya validado. No hay grupo de comparación y los autores declaran que el diseño no permite afirmar causas. Tampoco hay cifras de salarios ni de despidos. La banda de edad es angosta, y los autores advierten que quienes estén más cerca del retiro podrían sentir el efecto con más fuerza, algo que no se probó11.
Para entender por qué esto no es el desempleo de siempre, conviene ver qué hace esta tecnología. Las olas anteriores de automatización pegaron en tareas rutinarias, físicas o de reglas fijas. Los sistemas actuales de IA generativa apuntan a otro lado: al trabajo de cuello blanco, bien pagado, que valía por el juicio especializado de una persona12. No reemplaza la mano: reemplaza el criterio.
Los propios participantes describen ese paso con una precisión incómoda. No los corrieron: los dejaron ahí, con el mismo puesto y el mismo rango, aprobando lo que el sistema ya hizo. Una analista clínica sénior lo resumió: "la experta que no puede explicar del todo su propio análisis ya no es realmente la experta. Es la operadora"13. Lo curioso es que el daño no viene del despido, viene de la permanencia: los conservan para que una cara humana responda por la máquina.
Hay además una asimetría de arranque que cualquiera que haya migrado reconoce. "Un profesional alemán de mercadotecnia de mi edad nunca tuvo que caer al fondo y volver a subir… yo empiezo este reto de la IA ya agotada… Él está fresco para esta pelea. Yo estoy exhausta antes de empezar"14. La energía que se gastó en la primera prueba no se recupera para la segunda.
El silencio tiene su propia mecánica, y se entiende mejor mirando cómo se comportan las empresas. Una gerente de proyectos de tecnología contó que sus dudas sobre el gobierno del sistema fueron vistas como "resistencia al cambio" y releídas como un defecto de carácter [IT6]. Cuando el sesgo de siempre se disfraza de dato objetivo, ya no hay vocabulario para impugnarlo: "la IA no creó el sesgo. Le dio al sesgo viejo una razón moderna"15. Y por debajo de todo eso, muchos sostienen la fachada: "muchos de nosotros, inmigrantes de mediana edad, rehicimos nuestras carreras, vamos bien por fuera, con ese zumbido bajo por dentro… fácil de pasar por alto porque no nos estamos desmoronando"16.
Otros trabajos que se han hecho sobre este terreno ayudan a dimensionarlo, aunque solo se conoce su resumen; el texto completo no estaba disponible. Un estudio de Makridis y Hickman, según su resumen, lo único disponible de él, ligó requisitos de personalidad en el trabajo con más de 15 millones de trabajadores estadounidenses entre 2006 y 2024, y encontró que las exigencias de tenacidad intelectual se asociaron con salarios por hora más altos, mientras las de ajuste social tuvieron una asociación más débil y negativa1718. Esas tendencias empezaron antes de la IA generativa, pero se concentraron en ocupaciones más expuestas a ella19. Otro trabajo, de Arias y colegas, según su resumen, lo único disponible de él, distingue dos canales: el desplazamiento de trabajadores en ciertas tareas y la mejora de productividad en otras20. Y anota algo que le importa a esta región: en los países en desarrollo la adopción de IA todavía es incipiente y la de robots se limita a sectores de tareas manuales rutinarias y salarios altos21.
También hay investigación sobre el trabajo de plataformas, del que solo se conoce el resumen. Un estudio sobre Italia, según su resumen, lo único disponible de él, encontró que la vulnerabilidad económica ahí no se concentra en los más jóvenes, como suele suponerse, sino en personas mayores o más establecidas que enfrentan inestabilidad, subempleo o ingresos a la baja22. El trabajo de plataforma funcionaba como complemento de ingresos precarios23.
Así lo lee weeklyAI. El patrón que estos relatos dibujan no es el del trabajador que pierde el puesto, sino el del que conserva el sueldo y pierde la voz. Y eso tiene una consecuencia práctica: cuando el valor de alguien se mide por lo bien que defiende lo que produjo la máquina, el que solo sabe hacer bien el trabajo queda en desventaja frente al que sabe presentarlo. En hogares como los del lector, eso llevaría a que el reconocimiento visible no se traduzca en acceso: lo elogian, pero no lo meten a las conversaciones donde se reparte el aprendizaje de las herramientas nuevas. Esta lectura se equivocaría si los profesionales inmigrantes con buen desempeño entraran a los mismos círculos informales, mentores y cursos que sus pares nacidos en el país, y si su ascenso dependiera del trabajo y no del origen.
Hay un segundo hilo que vale la pena vigilar. Quien concentra la palabra en una empresa deja a los demás en un silencio que luego se lee como acuerdo. Es probable que las dudas técnicas sobre un sistema automático se digan solo en el pasillo o no se digan, y que la dirección tome esa calma como aceptación. El lector puede hacer algo concreto: si tiene reparos sobre una herramienta en su trabajo, déjelos por escrito por un canal que proteja su nombre, o reúnalos con compañeros antes de la junta, y pregunte cómo se van a tomar en cuenta las objeciones y quién las revisa. También puede notar qué premia su evaluación: el resultado técnico o la forma de presentarlo. Guarde evidencia de las decisiones en las que su criterio corrigió a la máquina.
Vuelva a la sala de lectura del hospital. La placa sigue ahí, la máquina ya opinó, y alguien tiene que firmar. Lo que estos 32 relatos dejan ver es que la pregunta útil no es si la IA puede hacer ese trabajo, sino quién queda con derecho a decir que no. ¿En su trabajo, quién firma, y a quién se le permite dudar?
De dónde sale cada dato de contexto, y cuánto leímos de cada documento
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Participants described the shift in consistent terms: from performing the work to approving the machine’s performance of it. One senior radiologist described how the clinical AI, framed as decision support, in fact read the images first while the clinician confirmed what it had found: “sometimes I feel I am not even asked to think anymore. I am asked to agree” (H1, Healthcare)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Artificial intelligence is rapidly disrupting high-level cognitive and analytical labour. Yet, while emerging macroeconomic analyses increasingly map AI exposure across broad demographic boundaries, there is a scarcity of qualitative research exploring how these shifts are experienced by highly skilled older immigrant professionals navigating the workforce at the compounding intersection of age, migration status, and professional isolation."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Ten themes revealed a pattern termed Double Displacement: the compounding of chronic socio-cultural exclusion from domestic professional networks with the acute techno-cognitive demotion from autonomous expert to passive machine validator."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "An emerging-markets analyst had spent 16 years building regional expertise in Eastern European financial markets, a knowledge edge rooted in origin that no locally born colleague could easily replicate. When AI platforms synthesised comparable coverage at scale, the advantage dissolved: “The machine took the good half of being foreign and left me the bad half” (F5, Finance)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A further finding, the foreignness premium, describes how AI systems expropriate the distinctive multicultural and regional professional assets that migration history conferred, turning these assets into drivers of late-career obsolescence."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A radiologist described his clinical output becoming “feedback to make the machine better… a data point for the machine’s training” (H1, Healthcare); a finance professional distilled the structure: “My expertise was treated as training data. As a reason to make the machine better so it does not need me next time” (F2, Finance)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A senior radiologist was told, informally, almost in a corridor, that AI upskilling programmes were being “prioritised for younger staff” (H1, Healthcare), and others received orientation sessions that amounted to a login and a link. The organisation that withholds reskilling access while simultaneously demanding AI fluency is not offering a proving ground. It is manufacturing a failure."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A finance professional named it without ambiguity: “the pension fear and the AI fear, they are the same fear, really. Both point at the same cliff” (H1, Healthcare). The compound is not coincidental. A displacement that in earlier career years would allow time to retrain and rebuild instead arrives at the precise moment when losing even a year of pension contributions is permanent."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Contributions split across two or more national systems left many with two incomplete halves, neither sufficient on its own, with no mechanism to combine them. A finance professional described the navigation as disorienting in itself: “a Belgian pension I do not understand and a Greek pension I don’t trust… The not-knowing is almost worse than bad news would be” (F7, Finance)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A senior clinical data analyst described the mechanism directly: “My silence becomes evidence that I had no problem. The system is built so that the people most affected cannot safely speak, and then their silence is used against them” (H2, Healthcare)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "The 50–55 age band is relatively narrow. Professionals in the 55–65 range, closer to the retirement threshold that Theme 6 identifies as a structural amplifier, may experience the compounding dynamic with greater intensity. Extending the age range in future research would test whether the harmful convergence becomes more or less acute as retirement approaches."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "Contemporary generative AI and machine learning systems follow a fundamentally different logic: they directly target cognitive, high-wage, white-collar occupations whose value has historically rested on specialised human judgement ( Lebovitz et al., 2021 )."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A senior clinical data analyst, now asked to validate outputs she had not generated, concluded that “the expert who cannot fully explain her own analysis is not really the expert anymore. She is the operator” (H2, Healthcare)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A marketing analyst described the structural asymmetry directly: “a German marketing professional my age never had to fall to the bottom and climb back… I’m starting this AI challenge already depleted… He′s fresh for this fight. I’m exhausted before it starts. Same fight, very different starting condition” (M3, Marketing)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A senior compliance officer described the dynamic: “AI did not create the bias. It gave the old bias a new, modern-looking reason. Now they can question me and it sounds like data-driven decision-making instead of prejudice” (F2, Finance)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - el artículo del que trata esta nota — el artículo completo — el pasaje: "A senior nurse gave this condition its name, observing that there were “a lot of us, middle-aged immigrants, rebuilt our careers, doing fine on the surface, with that low hum underneath… easy to miss because we are not falling apart” (H3, Healthcare)."
- Makridis C, Hickman L. (2026). Intellectual Tenacity at Work: Generative AI Exposure and the Economic Returns to Occupational Personality Requirements. Academy of Management Discoveries. https://doi.org/10.5465/amd.2025.0303 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "We link these measures to more than 15 million workers in the American Community Survey from 2006 to 2024."
- Makridis C, Hickman L. (2026). Intellectual Tenacity at Work: Generative AI Exposure and the Economic Returns to Occupational Personality Requirements. Academy of Management Discoveries. https://doi.org/10.5465/amd.2025.0303 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Intellectual tenacity requirements were positively associated with hourly wages, whereas social adjustment requirements had a weaker, negative association."
- Makridis C, Hickman L. (2026). Intellectual Tenacity at Work: Generative AI Exposure and the Economic Returns to Occupational Personality Requirements. Academy of Management Discoveries. https://doi.org/10.5465/amd.2025.0303 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "These divergent trends began before widespread GenAI adoption, but they are concentrated in occupations with greater exposure to GenAI: wage returns for intellectual tenacity requirements increased in occupations highly exposed to GenAI, while returns to social adjustment requirements decreased in those occupations."
- Arias O, Fukuzawa D, Le DT, Mattoo A. (2026). Jobs and Technology: An Integrated View. The World Bank Research Observer. https://doi.org/10.1093/wbro/lkag004 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Two channels of impact are also identified: displacement of workers in certain tasks, and productivity enhancement in other tasks."
- Arias O, Fukuzawa D, Le DT, Mattoo A. (2026). Jobs and Technology: An Integrated View. The World Bank Research Observer. https://doi.org/10.1093/wbro/lkag004 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "In developing countries, robot adoption has been limited to high-wage, routine manual task-intensive sectors, while AI adoption remains nascent."
- Punzi C, Cirillo V, Guarascio D, Pellungrini R, Giannotti F. (2026). Platform workers not by chance: Exploring the digital labour markets in Italy with machine learning and explainable AI. PLOS One. 10.1371/journal.pone.0340237 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Economic vulnerability is concentrated not among the youngest workers, as often suggested in the literature, but among older or more established individuals facing job instability, underemployment, or declining income from traditional occupations."
- Punzi C, Cirillo V, Guarascio D, Pellungrini R, Giannotti F. (2026). Platform workers not by chance: Exploring the digital labour markets in Italy with machine learning and explainable AI. PLOS One. 10.1371/journal.pone.0340237 — solo el resumen - no se pudo obtener el texto completo — el pasaje: "Moreover, the analysis reveals that platform work is associated with structural vulnerabilities typical of non-standard employment, including unstable contracts, gender inequalities, and economic insecurity, and it primarily functions as a compensatory mechanism to supplement insufficient earnings from precarious jobs."
Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614
Quién pagó: El trabajo fue financiado por la Agencia Eslovena de Investigación con la subvención P2-0037; el artículo no declara relaciones comerciales o financieras que constituyan conflicto de interés y no dice nada sobre el papel de los financiadores en el diseño ni sobre equipos o programas cedidos.
No tome esto como consejo médico profesional.
interviews · Frontiers in sociology · the paper, 4 Aug 2026 · free
"I Am Asked to Agree": When AI Leaves the Job and Takes the Judgement
In interviews with 32 foreign-born professionals aged 50 to 55 in four Western European sectors, several described checking a machine's work instead of doing their own. The study counted no wages and no layoffs, and cannot say AI caused any of it.
Short version · the longer version follows, about 7 min
Ask the weeklyAI chat
Ask me about AI in your life: I point you to the stories weeklyAI has published on it, this week and before.
Conversations are saved for as long as weeklyAI exists, to improve the publication. Answers come in the language you write in.
- The study at a glance
- Who
- Foreign-born professionals aged 50 to 55 in white-collar jobs
- How many
- 32
- Where
- Western Europe
- When
- not stated
- Kind of study
- interviews
- Who did it
- Jozef Stefan Institute, Ljubljana, Slovenia
- The limit that matters
- 32 interviews in four white-collar sectors; no pay or job-loss figures, no causal claims
The radiologist still reads the scan. What changed is who answers first.
"Sometimes I feel I am not even asked to think anymore. I am asked to agree," a radiologist said. One radiologist said the machine reads the scans first, and the specialist then confirms what it found.
Researchers interviewed 32 foreign-born professionals, ages 50 to 55, in healthcare, finance, marketing and information technology. Eight came from each sector, all living and working in Western Europe, and each interview ran about 55 minutes on average.
The pattern behind that sentence: people kept their jobs and often their rank. What changed was the work itself. They moved from making the judgement to approving output the machine had already produced. One IT project manager put it as going "from author to supervisor of the machine."
Some said their own expertise had trained the systems that then replaced them. A finance professional described his knowledge being used "as training data." Others said the regional knowledge and cross-cultural skills that migration had given them were automated away first. "The machine took the good half of being foreign and left me the bad half," one analyst said.
This is 32 interviews in four white-collar sectors. The study reports no measured figures on pay or job loss, and its authors say it makes no causal claims. It cannot tell you your own odds. The researchers call the pattern they named, Double Displacement, a framework awaiting testing.
One radiologist was told that AI training was being given to younger staff first. One IT project manager said her concerns were treated as resistance to change. Others were protected, they said, by the type of role they held, by timing, by networks, or by an employer that treated foreign-born staff as ordinary.
Watch for the validator shift in your own work. When your day starts to consist of approving output you did not produce, that is the moment the study points to: ask for training that actually reaches you, and find a way to raise concerns without paying for it.
What this means for you
What to watch for, then, is not a headline about robots taking jobs but the smaller shift inside a job that stays. If your work begins to look like approving what a system produced, and training goes to younger colleagues first, those are the things the interviewees described. This study cannot say it will happen to you.
Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614
Who paid: The work was funded by the Slovenian Research Agency under Grant P2-0037; the article states no commercial or financial relationships that could be a conflict of interest, and says nothing about funders' role in the design or about equipment or software lent.
Do not take this as professional medical advice.
The longer versionThe passages copied from the paper, the pictures, and every source with how much of it we read · 1378 words · about 7 minRead it →Close
Their Jobs Weren't Cut. The Machine Just Started Answering First.
Thirty-two foreign-born professionals aged 50 to 55, in four Western European knowledge-intensive sectors, described the same quiet demotion: from the person who decides to the person who approves what the system already decided.
Consider what a clinic visit now covers. A radiologist reads the scan, and a clinical AI has already read it first; the doctor's job is to confirm what the system found. One senior radiologist in the study put it plainly: "sometimes I feel I am not even asked to think anymore. I am asked to agree"1.
That is the shape of the finding, and it is not a layoff story. A senior clinical data analyst, asked to validate outputs she had not generated, said the expert who cannot fully explain her own analysis "is not really the expert anymore. She is the operator"2.
The study was qualitative. Martin Mihajlov and Tanja Pavleska, of the Jozef Stefan Institute in Ljubljana, Slovenia, interviewed 32 foreign-born professionals aged 50 to 55, eight each in healthcare, finance, marketing and information technology. All had lived and worked continuously in a Western European host country for at least 10 years. Interviews averaged 55.4 minutes, and participants also filled in two standard questionnaires on attitudes toward AI and on job insecurity. The researchers drew ten themes from the transcripts; the funding came from the Slovenian Research Agency, and the authors declared no commercial conflicts34.
The central pattern they named Double Displacement: years of being kept at a distance from domestic professional networks, then a sudden demotion from independent expert to checker of machine output4. A second finding they called the foreignness premium: the very assets migration gave these professionals — regional knowledge, fluency across cultures — were the first thing automated away5. One emerging-markets analyst had spent 16 years building expertise in Eastern European financial markets. When AI platforms produced comparable coverage at scale, the edge dissolved: "The machine took the good half of being foreign and left me the bad half"6.
The authors are careful about what they have. They call Double Displacement a framework "awaiting empirical test," and they say it stands whether or not the illustrative data represents actual participants. No causal claim is made: the design cannot show that AI caused a lost job, a smaller pension or a health problem. The age band is narrow, the authors note, and professionals aged 55 to 65, closer to retirement, might feel the squeeze more sharply7. No formal ethics review was required. Nothing here counts how many workers are affected.
The mechanism is worth slowing down for, because it runs in a direction most people don't expect. The knowledge these professionals accumulated did not simply become obsolete; it was used to train the systems that replaced them. A radiologist described his clinical output becoming "feedback to make the machine better… a data point for the machine's training"8. A finance professional said it flatly: "My expertise was treated as training data. As a reason to make the machine better so it does not need me next time"8.
What kind of technology does this? The study's own framing is that today's generative AI and machine learning systems break from earlier automation: they go after cognitive, high-wage, white-collar work whose value rested on specialised human judgement, rather than the routine and physical tasks earlier waves hit9. The authors note that qualitative research on how this lands on older immigrant professionals is scarce, even as broad economic analyses map AI exposure across populations3.
The pension arithmetic is where the two fears merge. For a worker in the final decade, a year without contributions is a year that does not come back. One participant said the pension fear and the AI fear "are the same fear, really. Both point at the same cliff"10. For the foreign-born there is an extra turn of the screw: contributions split across two or more national systems can leave two incomplete halves, neither sufficient alone, with no mechanism to combine them. "A Belgian pension I do not understand and a Greek pension I don't trust," one finance professional said11.
Training is the other hinge. A senior radiologist was told informally, almost in a corridor, that AI upskilling was being "prioritised for younger staff," while others got orientation sessions amounting to a login and a link12. And the starting line was not level to begin with: a marketing analyst compared herself to a German colleague her age — "He's fresh for this fight. I'm exhausted before it starts. Same fight, very different starting condition"13.
Silence does the rest. A senior clinical data analyst described how it works: "My silence becomes evidence that I had no problem. The system is built so that the people most affected cannot safely speak, and then their silence is used against them"14. A senior nurse named the daily cost: many middle-aged immigrants, careers rebuilt, "doing fine on the surface, with that low hum underneath… easy to miss because we are not falling apart"15. And when bias enters the picture, it arrives wearing new clothes: "AI did not create the bias. It gave the old bias a new, modern-looking reason," a compliance officer said16.
Here is weeklyAI's reading. The article is not really about machines getting better at tasks. It is about who is allowed into the room where the tool is taught and who keeps the authority to overrule it. Expect the pattern to show up wherever a workplace adopts a system and hands out training unevenly: the people kept formally employed but pushed out of the learning loop will be judged on how well they defer. This reading would be wrong if workers shut out of reskilling, with no standing to challenge the machine, routinely kept their old judgement and authority anyway. What you can do with it is concrete: ask who is on the training list and who decides what the tool may overrule, and get the answer in writing.
A second thread weeklyAI draws from these accounts: people first stop trusting their own read of a situation, then stop saying so, and the quiet gets counted as agreement. The study's participants described both, and neither is a personal failing. The article recommends that organisations set up anonymous reporting channels and protected ways to raise dissent, so that people can speak without fear. When a rollout produces unanimous enthusiasm, ask a colleague privately what they actually think before treating the silence as consent.
The findings come with their own fence. The authors say the results should not be generalised beyond knowledge-intensive sectors with active AI integration, and the sectors were picked precisely because adoption there is advanced. They also note buffer conditions: some participants were protected by role type, employer culture or pension position, and one described her comfort as "situational, not earned… it's about where I am, not who I am"5. Protection, where it existed, was real and time-limited.
Other work puts this in a wider frame. According to the summary of a study of more than 15 million American workers, the only part of it available; the full paper is behind a subscription, wage returns for persistence and analytical effort rose after 2014 and were concentrated in occupations highly exposed to generative AI, while returns to social-adjustment traits declined171819. According to the summary of a review of jobs and technology, the only part available, AI threatens fewer jobs in developing countries because a smaller share of employment sits in cognitive tasks, though those countries are also less equipped to benefit from it20. According to the summary of a study of platform work in Italy, the only part available, economic vulnerability there was concentrated not among the youngest workers but among older or more established people facing job instability or declining income21.
This reading would be wrong if older skilled workers in the reader's own country were routinely included in retraining on the same terms as younger colleagues and still ended up as machine checkers. What to watch for: whether your employer is teaching you to run the system or only to approve it.
The radiologist still reads the scan. What changed is who answers first.
Ask, in writing: who is on the training list, and who decides what the machine may overrule?
Where each piece of context comes from, and how much of it we read
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "Participants described the shift in consistent terms: from performing the work to approving the machine’s performance of it. One senior radiologist described how the clinical AI, framed as decision support, in fact read the images first while the clinician confirmed what it had found: “sometimes I feel I am not even asked to think anymore. I am asked to agree” (H1, Healthcare)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "A senior clinical data analyst, now asked to validate outputs she had not generated, concluded that “the expert who cannot fully explain her own analysis is not really the expert anymore. She is the operator” (H2, Healthcare)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "Artificial intelligence is rapidly disrupting high-level cognitive and analytical labour. Yet, while emerging macroeconomic analyses increasingly map AI exposure across broad demographic boundaries, there is a scarcity of qualitative research exploring how these shifts are experienced by highly skilled older immigrant professionals navigating the workforce at the compounding intersection of age, migration status, and professional isolation."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "Ten themes revealed a pattern termed Double Displacement: the compounding of chronic socio-cultural exclusion from domestic professional networks with the acute techno-cognitive demotion from autonomous expert to passive machine validator."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "A further finding, the foreignness premium, describes how AI systems expropriate the distinctive multicultural and regional professional assets that migration history conferred, turning these assets into drivers of late-career obsolescence."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "An emerging-markets analyst had spent 16 years building regional expertise in Eastern European financial markets, a knowledge edge rooted in origin that no locally born colleague could easily replicate. When AI platforms synthesised comparable coverage at scale, the advantage dissolved: “The machine took the good half of being foreign and left me the bad half” (F5, Finance)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "The 50–55 age band is relatively narrow. Professionals in the 55–65 range, closer to the retirement threshold that Theme 6 identifies as a structural amplifier, may experience the compounding dynamic with greater intensity. Extending the age range in future research would test whether the harmful convergence becomes more or less acute as retirement approaches."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "A radiologist described his clinical output becoming “feedback to make the machine better… a data point for the machine’s training” (H1, Healthcare); a finance professional distilled the structure: “My expertise was treated as training data. As a reason to make the machine better so it does not need me next time” (F2, Finance)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "Contemporary generative AI and machine learning systems follow a fundamentally different logic: they directly target cognitive, high-wage, white-collar occupations whose value has historically rested on specialised human judgement ( Lebovitz et al., 2021 )."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "A finance professional named it without ambiguity: “the pension fear and the AI fear, they are the same fear, really. Both point at the same cliff” (H1, Healthcare). The compound is not coincidental. A displacement that in earlier career years would allow time to retrain and rebuild instead arrives at the precise moment when losing even a year of pension contributions is permanent."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "Contributions split across two or more national systems left many with two incomplete halves, neither sufficient on its own, with no mechanism to combine them. A finance professional described the navigation as disorienting in itself: “a Belgian pension I do not understand and a Greek pension I don’t trust… The not-knowing is almost worse than bad news would be” (F7, Finance)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "A senior radiologist was told, informally, almost in a corridor, that AI upskilling programmes were being “prioritised for younger staff” (H1, Healthcare), and others received orientation sessions that amounted to a login and a link. The organisation that withholds reskilling access while simultaneously demanding AI fluency is not offering a proving ground. It is manufacturing a failure."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "A marketing analyst described the structural asymmetry directly: “a German marketing professional my age never had to fall to the bottom and climb back… I’m starting this AI challenge already depleted… He′s fresh for this fight. I’m exhausted before it starts. Same fight, very different starting condition” (M3, Marketing)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "A senior clinical data analyst described the mechanism directly: “My silence becomes evidence that I had no problem. The system is built so that the people most affected cannot safely speak, and then their silence is used against them” (H2, Healthcare)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "A senior nurse gave this condition its name, observing that there were “a lot of us, middle-aged immigrants, rebuilt our careers, doing fine on the surface, with that low hum underneath… easy to miss because we are not falling apart” (H3, Healthcare)."
- Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614 - the article this story is about — the whole article — the passage: "A senior compliance officer described the dynamic: “AI did not create the bias. It gave the old bias a new, modern-looking reason. Now they can question me and it sounds like data-driven decision-making instead of prejudice” (F2, Finance)."
- Makridis C, Hickman L. (2026). Intellectual Tenacity at Work: Generative AI Exposure and the Economic Returns to Occupational Personality Requirements. Academy of Management Discoveries. https://doi.org/10.5465/amd.2025.0303 — only the abstract - the full text could not be fetched — the passage: "We link these measures to more than 15 million workers in the American Community Survey from 2006 to 2024."
- Makridis C, Hickman L. (2026). Intellectual Tenacity at Work: Generative AI Exposure and the Economic Returns to Occupational Personality Requirements. Academy of Management Discoveries. https://doi.org/10.5465/amd.2025.0303 — only the abstract - the full text could not be fetched — the passage: "The wage premium for intellectual tenacity requirements was approximately stable through 2014 and rose thereafter, while the return to social adjustment requirements declined."
- Makridis C, Hickman L. (2026). Intellectual Tenacity at Work: Generative AI Exposure and the Economic Returns to Occupational Personality Requirements. Academy of Management Discoveries. https://doi.org/10.5465/amd.2025.0303 — only the abstract - the full text could not be fetched — the passage: "These divergent trends began before widespread GenAI adoption, but they are concentrated in occupations with greater exposure to GenAI: wage returns for intellectual tenacity requirements increased in occupations highly exposed to GenAI, while returns to social adjustment requirements decreased in those occupations."
- Arias O, Fukuzawa D, Le DT, Mattoo A. (2026). Jobs and Technology: An Integrated View. The World Bank Research Observer. https://doi.org/10.1093/wbro/lkag004 — only the abstract - the full text could not be fetched — the passage: "AI threatens fewer jobs in developing countries because of the relatively low share of employment in cognitive tasks, but these countries are also less equipped to benefit from AI."
- Punzi C, Cirillo V, Guarascio D, Pellungrini R, Giannotti F. (2026). Platform workers not by chance: Exploring the digital labour markets in Italy with machine learning and explainable AI. PLOS One. 10.1371/journal.pone.0340237 — only the abstract - the full text could not be fetched — the passage: "Economic vulnerability is concentrated not among the youngest workers, as often suggested in the literature, but among older or more established individuals facing job instability, underemployment, or declining income from traditional occupations."
Mihajlov, M., Pavleska, T. (2026). Too old, too foreign, too replaceable? How AI shapes livelihood security of ageing immigrants. Frontiers in Sociology. https://doi.org/10.3389/fsoc.2026.1897614
Who paid: The work was funded by the Slovenian Research Agency under Grant P2-0037; the article states no commercial or financial relationships that could be a conflict of interest, and says nothing about funders' role in the design or about equipment or software lent.
Do not take this as professional medical advice.
