Back to projectsBackground Information
The International Labour Organization estimates that there are 4.8 million people being trafficked around the world for commercial sex, a global industry estimated to be worth $99 billion USD (ILO and Walk Free, 2017, ILO, 2014). Sex trafficking accounts for around 19.3% of trafficking victims worldwide (ILO, 2014). The United Nations Office on Drugs and Crime highlighted that detected cases of sex trafficking are disproportionately prevalent in North America (UNODC, 2020).
Technology has been a critical tool for traffickers to recruit, advertise and exploit victims, all the while making their activity more elusive and widespread (UNODC, 2020). Technology is often used for advertising purposes, with a majority of detected victims being advertised online. In the past two decades, 81% of all prosecuted sex trafficking cases involved solicitations online (Trafficking Institute, 2020). Analyzing suspicious activity online can allow us to identify and support victims of human trafficking as well as understand the prevalence and patterns of sex trafficking online. This is especially important since current statistics on sex trafficking offer only a rough estimate of the number of human trafficking victims (both globally and in Canada), given that relevent data is collected in non-standardized, siloed and unreliable ways (Szablewska et al., 2018, Russell, 2017, Farrell, 2017, Weitzer, 2015, Goodey, 2008).
Project Description
Traffickers are increasingly leveraging social media, classified online advertising and forums to advertise human trafficking victims (Sarker, 2015). The goal of this project is to help design AI solutions to sift through various types of advertisements and flag those that are likely to contain victims of organized human trafficking.
Our designed techniques are at the intersection of pattern detection, data mining, active and human-in-the-loop learning, anomaly detection, graph mining, natural language processing, information retrieval, and image processing.
Mila is a member of Code 8.7, a community that develops and applies AI-powered and survivor-informed anti-slavery solutions.
Relevance to Canada
The Canadian landscape has been identified as a source, transit and destination (or seller) country for human trafficking, predominantly sex trafficking (Public Safety Canada, 2019). A report that analyzed over a thousand charges related to trafficking in persons in Canada found that 85% of charges came from Ontario and Quebec, occurring primarily within urban centers (Millar and O’Doherty, 2020). The majority of these victims have been women and girls. The number of charges and convictions in Canada may not be reflective of the number of incidents but may be more indicative of the criminal justice system’s challenge in properly identifying and prosecuting human trafficking cases (O’Doherty, 2018).
In the Canadian context, innovative technological solutions designed to help fight human trafficking are lacking. It is the mission of our interdisciplinary team, composed of both AI and criminology researchers, to find ways of filling this gap. We understand the complexities of this domain and seek to accomplish our goal responsibly by engaging with the right stakeholders and conducting context-aware, human-centered, and human-rights-based research.
Declaration
Statement for the Ethical Development of AI to Counter Trafficking
Mila is committed to the responsible development and use of AI systems. This commitment informs our approach on this project, which grounds our work in human-centric governance principles.
The research team leading this project is both comprised of and engaged with those who are acutely aware of the risks and considerations of this work, particularly as it relates to the problematic misidentification of sex workers as sex trafficking victims.
Although our research project has not been deployed, we are continuously implementing harm prevention procedures such that our solutions will not be able to be misused or create harm to vulnerable communities, specifically Indigenous peoples, immigrant/non-status peoples, members of the LGTBQ2S+ community and sex workers. Our team has already undertaken significant efforts to understand these complexities and is working closely with domain experts as well as a survivor of trafficking in the design of our project to ensure ethical/inclusive stakeholder engagement. Moreover, the team is pursuing consultations with marginalized communities, sex worker rights organizations and survivors of trafficking to ensure their needs and perspectives are reflected meaningfully in the project.
The project is currently at the research stage. If and when this work becomes operational, the technology being developed would be managed independently, with a dedicated manager to oversee and ensure it is being used as intended. The research team will continue working with a dedicated group of lawyers and research ethicists to ensure this work abides by local privacy and security laws. This research project is not affiliated with Mila partners. Furthermore, all of the data collected is protected by our ethics protocol and is only accessible to certain members of the research team.
It is also important for us to acknowledge the existence of exaggerated claims and false narratives surrounding the human trafficking industry and concentrate instead on the realities of those facing exploitation in the commercial sex industry (see Miller & O’Doherty 2020, Sayers, 2018, Weitzer 2005). Scholars suggest being critical of official statistics as they often represent the priorities and policies of specific stakeholders rather than the realities of human trafficking in Canada (Millar & O’Doherty, 2020). Furthermore, statistics on sex trafficking may be more representative of a global preoccupation with exploitation in the commercial sex industry rather than with giving an accurate depiction of what is taking place(Morcom & Scholenhardt, 2011).
Frequently Asked Questions
1. What phase of development is the project in currently?
For the last several years, our researchers have been attempting to establish the technical feasibility of algorithms that are useful in the anti-human trafficking context. Now that the efficacy of those algorithms has been demonstrated, we are bringing those algorithms out of abstraction and researching their utility in a more applied way.
We’ve assembled a strong team of criminologists to ensure that the project is robust in addressing both the opportunities and risks of working within the human trafficking context.
As we undertake this work, we’re doing our best to ensure it is conducted with as narrow a goal as possible in order to facilitate victim-centric anti-human trafficking support. The reason for this focus is to prevent our work from implicating any other group of people. As such, we’re speaking with relevant stakeholders to see what constraints are needed to maximize the potential for meaningful intervention and reduce the risk of unintended harm.
2. What are the potential benefits of this research?
We plan to publish our research within both criminology and machine learning publications. Hopefully, this will inspire future work on the subject to be similarly multidisciplinary and rigorous.
In the long term, this research may be used to support service providers by providing them with access to targeted human trafficking-specific insight.
3. Do you have specific funding for this project?
The research carried out to create our algorithms has been financially supported by funding from public research funds as well as a grant from Mila, our academic research institute.
4. Who do I contact to learn more about this work?
If you’d like to get in touch, please reach out to Allison Cohen at allison.cohen@mila.quebec.
This research project is conducted under the leadership of Professor Reihaneh Rabbany, with the collaboration of multiple AI professors and students as well as criminology students specialized in the field of human trafficking, sex work and critical perspectives in the domain. For any inquiry about the project, please communicate with Allison Cohen, Mila’s AI for Humanity Applied AI Projects Lead.