PhD FindDoctoral opportunities · PhDFind
Menu
中文

Responsible AI for Online Safety PhD: Manipulation, Social Bots and Scams

Project 2714 · Leiden University · Faculty of Science; Leiden Institute of Advanced Computer Science (LIACS), System and Security cluster

Application workspace

Save this opportunity and prepare around its actual requirements. Confirm deadlines and eligibility at the original source.

Read the research report View the recruitment source
Country or region
Netherlands
Application deadline
2026-12-13 (confirm the exact time with the source)
Funding
Full-time employed PhD, 38 hours/week, €3,204–€4,051 gross/month, plus 8% holiday allowance, 8.3% year-end bonus and ABP pension. One-year contract with a conditional three-year extension.

The research opportunity

Project overview

Leiden University recruits a full-time PhD candidate to conduct methodological and empirical research on responsible AI for online safety. Possible directions include detecting AI-driven manipulation and social bots, early warning for fraud and scams, and human–AI responses to manipulation. The advert does not fix a single thesis topic or promise a dataset or platform access. Work includes publishing, presenting, teaching assistance and a doctoral thesis.

Research environment

The position is based in LIACS's System and Security cluster and connected with the Leiden Organizational Cyber Security Group. Dr Valerio La Gatta is the daily supervisor and Dr Olga Gadyatskaya co-supervises. The candidate will work with LIACS colleagues and international partners, whose identities and specific contributions are not stated. Hybrid work within the Netherlands may be possible when the work arrangements permit. Employment and Graduate School admission are separate steps; confirm registration, the training plan, data access and ethics arrangements.

Applicant requirements

Applicants need an MSc or equivalent in computer science, AI, data science or a related quantitative field, a solid machine-learning background, programming skills such as Python and experience with real-world data. The advert also asks for interest in societal AI and interdisciplinary research, independent and collaborative working, clear communication and English proficiency. NLP, large language models, network analysis and previous research are preferred, not mandatory; Dutch is not required.

Pay and application

The 38-hour-per-week position pays €3,204–€4,051 gross per month full-time, plus 8% holiday allowance, 8.3% year-end bonus and ABP pension. It starts with a one-year contract and can be extended by three years following a positive assessment of thesis progress, personal capabilities and mutual fit; extension is not automatic. The expected start is March 2027 or as soon as possible thereafter.

Apply by 13 December 2026 at 23:59 CET. Submit a one-to-two-page cover letter, CV, bachelor's and master's transcripts, and master's diploma or expected completion date. A writing sample and code/software links are optional; provide up to two references, contacted only for shortlisted candidates. Ask about the likely research direction, dataset access, supervision, teaching load and Graduate School process. Contact Dr Valerio La Gatta at v.la.gatta@liacs.leidenuniv.nl; application enquiries: jobs@liacs.leidenuniv.nl.

Research and application brief

Report generated: . This date does not confirm that recruitment remains open.

Information updated: . Recheck the original notice before applying.

Show report contents

Responsible AI for Online Safety PhD: Manipulation, Social Bots and Scams

Quick Decision

This is a full-time, paid doctoral-research appointment in responsible AI and online safety, based in Leiden University’s computer-science environment. It is most directly aligned with applicants who can already build and evaluate machine-learning systems on real-world data and want to study online manipulation, social bots, fraud/scams, or human–AI responses. The advertised work is both methodological and empirical; the vacancy names several possible directions rather than fixing a thesis question. The main practical condition is a one-year initial contract, with a further three years only after a positive progress and performance review. The deadline is 13 December 2026 at 23:59 CET. Before applying, prepare evidence of applied ML/programming and ask which research direction and data access are currently most likely. Official vacancy

Research Problem and Doctoral Work

What Is Unsolved and Why It Matters

The project addresses how AI can help protect online spaces while online manipulation itself becomes more automated and adaptive. The advertisement gives three illustrative, non-exclusive directions: modeling or detecting AI-driven manipulation such as coordinated influence campaigns and social bots; developing early-warning approaches for online harms such as fraud and scams; and studying how people and AI can jointly recognize and counter manipulation. These are stated as possible topics, not a commitment to all three or to a specific platform or threat dataset. Official vacancy

Recent work illustrates why the problem is technically and socially difficult. A 2026 preprint compares human, AI and hybrid detection of adaptive reinforcement-learning social bots in a controlled five-day study; because it is a preprint and a controlled experiment, it does not establish field performance. Adaptive social-bot study A 2025 paper examines positive and negative knowledge transfer across disinformation-detection tasks, highlighting that multi-task learning can help one task while harming another. Task-transfer paper A 2023 conference paper retrieves previously fact-checked claims to help identify false or unsupported claims in posts during the Russia–Ukraine conflict. Its narrow event and platform scope should not be generalized to scams or other harms. Claim-retrieval paper

The regulatory setting is also moving. The EU Digital Services Act identifies coordinated manipulation, fake accounts and bots among possible systemic risks for very large platforms, while Commission guidance and reporting describe platform risk-assessment duties and current implementation. These rules provide context for research questions, but do not prescribe this PhD’s methods or guarantee access to platform data. Digital Services Act, Commission implementation guidance, Commission systemic-risk report

Work, Methods, and Potential Outputs

Advertised tasks: conduct original methodological and empirical research; publish and present at leading venues in AI, computational social science and cybersecurity; contribute to teaching as a teaching assistant or head teaching assistant; and complete a PhD thesis. The vacancy does not name a required model family, dataset, platform, experimental design, or ethics protocol. Official vacancy

Evidence-informed preparation, not fixed duties: the cited work suggests useful questions about robustness as manipulation adapts, transfer across detection tasks, retrieval of corroborating evidence, and the benefits and failure modes of human–AI collaboration. A 2023 paper on interconnected online harms and moderation offers a broader framing, while a 2025 chapter on generative-AI agents in social-network simulations is relevant only if simulation becomes part of the eventual design; simulation validity would need independent testing. Online-harm paper, social-network simulation chapter

The advertisement’s preferred topics should not be mistaken for confirmed resources. It does not state what data or partner access exists, whether research will use public or partner data, how harm will be operationalized, or how model performance and human collaboration will be evaluated. The EU AI Act is a risk-based regulatory framework; it does not itself classify every online-safety detector as high-risk or define the project’s evaluation protocol. EU AI Act

Field and career context

The saved ENISA assessment describes European cybersecurity skills and recruitment gaps. This is adjacent labour-market context, not a forecast of Dutch online-safety PhD vacancies or an employment guarantee for this project. ENISA NIS Investments 2025

School, Programme and Field Context

The employer and research setting is Leiden University’s Faculty of Science, within the Leiden Institute of Advanced Computer Science (LIACS), System and Security cluster. The vacancy names Dr. Valerio La Gatta as daily supervisor and Dr. Olga Gadyatskaya as co-supervisor; both are connected with the Leiden Organizational Cyber Security Group (LOCS). LIACS describes a multidisciplinary computer-science institute, and LOCS works across organizational cybersecurity, usable security and technical security. These institutional pages establish research context, not guaranteed project resources or a particular supervision cadence. LIACS, LOCS, LOCS publications, Olga Gadyatskaya’s academic profile

QS places Leiden University at #=119 in its 2027 overall world ranking. This is institution-wide context, not a computer-science subject ranking or a measure of this research group. LIACS’s public research self-evaluation covers 2015–2020 and offers dated institutional context; it is not a current league table. QS 2027 overall ranking, LIACS research self-evaluation

The doctorate is governed by Leiden University regulations. The University’s 2024 regulations describe the route from an advertised paid PhD position through supervisor appointment and Graduate School admission; the PhD track starts once the Graduate School admits the candidate. University guidance describes training and supervision planning, and a trajectory diagram shows an approximate four-year route with a 9–12-month go/no-go review. Older guidance specifies 140 hours each for academic and transferable-skills training for employee PhDs; confirm how that guidance is currently applied by the relevant Graduate School. The vacancy is an employment appointment, while doctoral registration and progression remain subject to the University’s rules. PhD Regulations 2024, PhD Guidelines, PhD trajectory

Supervisor and Lab Assessment

Supervisor Research Directions

The vacancy places the work in the System and Security cluster and LOCS. La Gatta’s recent work spans social-bot detection, disinformation detection and information retrieval; Gadyatskaya’s profile and group publications show interests in organizational cybersecurity, risk management and secure software. This overlap can support a project that combines computational detection with security and human factors, but the exact division of supervision and each supervisor’s time commitment are not specified. LOCS, LOCS publications, Gadyatskaya profile

Recent Important Publications

  • La Gatta and colleagues’ 2025 Information Sciences paper studies positive and negative transfer between single-task and multi-task disinformation detection. It is useful when considering generalization across detection tasks, but is not a prescribed method for this project. Paper
  • La Gatta and colleagues’ 2023 ACM Web Conference Companion paper retrieves previously fact-checked claims from social posts. It provides an example of evidence retrieval in a specific conflict setting, not proof of effectiveness for fraud detection. Paper
  • La Gatta and colleagues’ 2026 arXiv preprint compares human, AI and hybrid approaches to detecting adaptive social bots. It is particularly close to the advertised human–AI and manipulation themes, but remains a preprint and its controlled five-day setup limits real-world inference. Preprint
  • Gadyatskaya and co-authors’ 2025 ACM study surveys 105 developers on perceived security impacts of generative-AI coding tools. It is adjacent evidence about human trust and AI security rather than online-content moderation. Paper

Lab, Resources, and Collaboration Environment

LOCS describes work on cybersecurity tools, explainable AI, security automation, organizational processes and people. The vacancy also mentions collaboration with LIACS colleagues and international partners, but does not name those partners or promise access to their systems or data. Ask what infrastructure, datasets, partner roles and external collaboration are actually available to this project. LOCS, Official vacancy

Funding, Application, and Living Cost

The appointment is full-time at 38 hours per week, initially for one year. A further three years is conditional on positive assessment of thesis progress, personal capabilities and compatibility. Expected start is March 2027 or as soon as possible thereafter. The gross monthly salary is €3,204–€4,051 at full-time, with 8% holiday allowance, an 8.3% year-end bonus, ABP pension, public-transport commuting reimbursement and at least 29 days of leave. The final salary step and individual net pay are not stated. Official vacancy, Dutch university collective agreement, 2026 salary scales, employment conditions

The University’s 2026–27 student budget estimate is €1,020–€2,020 per month. Treat this only as a student-cost proxy: employee taxes, rent, insurance and personal circumstances differ. Staff and PhD candidates can register with the University housing service, but availability is limited and the University advises starting early; registration is not a room guarantee. Build a personal budget and a fallback housing plan before accepting. University student budget, staff/PhD housing portal, housing search guidance

Applications are due by 13 December 2026 at 23:59 CET; applications received before the deadline are considered equally, and early applications are welcomed. Prepare a one-to-two-page cover letter, CV, BSc and MSc transcripts, and the MSc diploma or expected completion date. A writing sample and code/software links are optional; provide up to two reference contacts, who are contacted only for shortlisted candidates. Official vacancy

Applicant Fit and Preparation

Required Background and Skills

The vacancy requires an MSc or equivalent in computer science, AI, data science, or a related quantitative field such as mathematics, physics or computational social science; a solid machine-learning background; strong programming (Python/PyTorch are examples); and experience with real-world data. It also asks for an interest in societal AI and interdisciplinary research, independence and teamwork, clear communication, and English proficiency. Dutch is not required. These are advertised conditions, not a guarantee that a specific degree title will be judged equivalent. Official vacancy

Preferred Experience and Suggested Preparation

Preferred, not mandatory: NLP, large language models, network analysis and prior research experience such as a research thesis or publication. Suggested preparation: show one or two concrete projects where you handled noisy or shifting data, established a baseline, evaluated error types and explained limitations. If proposing a human–AI study, describe how you would measure assistance and potential overreliance without assuming that this is the project’s chosen method. A 2025 simulation chapter and the EU regulatory material can help frame questions, but neither creates an advertised skill requirement. Official vacancy, simulation chapter, EU AI Act

The position may be less aligned with applicants seeking a predetermined narrow thesis, guaranteed external dataset access, or a fully specified human-subject protocol before joining. Those features are not described in the vacancy. This is a scope limitation, not an eligibility rule.

Next Actions and Questions to Verify

Before Applying

  1. Tailor the cover letter to one or two advertised directions and show evidence of machine-learning implementation on real-world data.
  2. Confirm that the degree and transcript meet the quantitative MSc/equivalency requirement.
  3. Ask which direction is currently most likely, whether a research proposal is expected, and what the Graduate School admission sequence requires.

At Interview

  1. Which datasets and platform or industry partners are available, and what are the access, privacy and sharing limits?
  2. How will harms be defined and model performance evaluated? What human-subject ethics review may be needed?
  3. How are research, publications and teaching assistance balanced within the 38-hour week?
  4. What milestones and review criteria determine the extension after year one, and how often will each supervisor meet the candidate?

Before Accepting an Offer

  1. Which salary step applies, and how are allowances, pension and commuting reimbursement administered?
  2. What Graduate School registration and training obligations apply to this employee PhD, including current implementation of training-hour guidance?
  3. What housing support is realistically available, and what is the budget if University housing is unavailable?

Evidence and Sources

Project, university, programme, and supervisors

Research publications and policy context

Employment and cost

Sources

The listing may come from a university or recruitment platform. Check the notice and official sources in the report for current application requirements. View the recruitment source

Explore opportunities with related research or the same country, prioritising similar funding where relevant. Recheck each notice before applying.

Leiden University · Netherlands

Indigenous Heritage PhD | Mapuche Histories and Museum Collections

NWO-funded full-time employment: one year, extendable for three years after positive evaluation; €3,204–€4,051 gross monthly plus 8% holiday allowance and 8.3% year-end bonus. Tax-facility eligibility is individual.

2027-01-01 (confirm the exact time with the source)

Universitair Medisch Centrum Utrecht (UMC Utrecht) · Netherlands

PhD in Healthcare Design: Designing Patient-Centred Self-Service MRI Pathways in Primary Care

UMC Utrecht lists gross OIO pay of EUR 3,217–4,077/month on a 36-hour basis; this role is 32–36 hours/week, while another project-record line calls EUR 3,217 the maximum. Confirm the scale, step and hours-based pay with HR. Listed benefits: 8.3% year-end, 8% holiday, a 70% employer ABP pension contribution and commuting support. The Reimagine Imaging programme total is not a personal stipend. Contract term, start date, net pay and individual funding are unstated.

2026-11-17 (confirm the exact time with the source)

Eindhoven University of Technology (TU/e) · Netherlands

PhD in Random Networks: How Does Network Structure Affect Phase Transitions in Spreading and Consensus Processes?

Three full-time salaried PhD positions for four years; EUR 3,204–4,051 gross/month (Dutch university scale P), plus 8% holiday allowance and 8.3% year-end bonus, pension and other employee benefits. International staff immigration support is offered; the expat tax compensation scheme is subject to individual eligibility and its value must be confirmed.

2026-12-02 (confirm the exact time with the source)