Reporting sexual assault is a very sensitive and nuanced matter, and identifying at-risk groups may illuminate areas in which rape and sexual assault policy needs to improve.
Drawing on a large database of publicly available criminal reports from two US cities, a team led by Konstantin Klemmer of the University of Warwick, UK, and New York University, US, used a machine-learning approach to assess common trends in delayed reporting of rape.
Fast Facts: Sexual Assault
- Sexual assault in a major health and welfare problem in Australia
- Sexual assault of females is 7 times higher than males
- 1 in 3 hospitalised sexual assault cases are caused by a domestic partner
- Nearly 2 million Australia have experienced sexual assault since age 15
Published in The Royal Society Open Science, the study reveals that delayed reporting was common for younger survivors, for those in low socio-economic areas, and during holiday seasons.
Klemmer explains that understanding these trends is important for multiple reasons, including identification of vulnerable groups to provide better medical and mental health support.
“Eventually, we hope that our study can help to motivate data-driven policies and public-sector interventions targeted at exactly these vulnerable groups,” he says.
University of Melbourne criminologist Bianca Fileborn, who was not involved in the study, echoes that sentiment. “In my view, the biggest benefit of this type of spatial analysis is that it can be used to inform policy making and target resources towards survivors in locations that are currently under-serviced and under-resourced,” she says.
“Having access to appropriate rape crises counselling and ongoing therapeutic support, for example, can be vital for survivors’ healing.”
Already, only 1 in 4 rapes are reported in America, and only 1 in 10 women in Australia seek police support, but these proportions are likely not representative considering the number of people who do not disclose a sexual assault.
“Rates of under-reporting vary somewhat across countries, but this seems to be a fairly universal occurrence,” says Fileborn.
Previously, data on delayed rape reporting was undertaken through one-on-one interviews, which doesn’t necessarily reveal minute trends at a population level.
“Machine learning is great at finding patterns in large and noisy datasets. As such, it is a useful tool for finding associations within the complex rape reporting database we constructed,” says Klemmer.
“These datasets are rather new, large and complex and our study is the first to tap this data source.
“However, it is important to note its limitations here: while machine learning can be very successful at finding patterns, it isn’t necessarily great at explaining these patterns and finding causal relationships.”
This means that, while the data from this study is an important step towards identifying patterns, the actual why of the question requires a different type of data that “can really only be achieved through qualitative approaches that ask survivors about their experiences and reasons for non or delayed reporting,” according to Fileborn.
“There are some benefits to using this [machine learning] type of approach – for example, it reduces the need to have survivors re-tell their experiences, which can also be upsetting and retraumatising.
“However, I think this type of approach needs to be used in combination with survivor-centred research in order to fully understand the complexity and nuance of non/delayed reporting.”
Ultimately, the complex decisions around rape reporting are very individual and often driven by personal and social pressure. Fileborn points out that many survivors experience a deep sense of shame and self-blame and feel that what happened was too ‘minor’ or ‘trivial’ to report. Many also believe that nothing can be done in response to what happened, or fear retaliation from their assaulter.
This means that decisions about whether to report a sexual assault can be compared to the trends found by machine learning, to see if there are some areas that haven’t been researched enough. Comparing the two bits of data may help establish why these trends are happening.
“It was interesting to see that Federal holidays were associated with delayed reporting,” remarks Fileborn. “This might be related to people consuming alcohol and partying, factors that are associated with sexual violence occurring, but that are also used to blame survivors for their own victimisation – i.e., because ‘they drank too much’.
“Such findings might point to the need to continue to educate the community and tackle myths about the role that alcohol plays in sexual violence.”
Using machine learning distils Kremmer’s so-called “noisy” data sets, but anonymity is of utmost importance. Why does anonymity matter?
“This is a super important question!” says Klemmer. “Data on sexual violence is extremely sensitive and publishing it online must be done with special caution in order to protect survivors.”
This data, in conjunction with personal descriptions or decisions, could hopefully isolate some of the causes of non-reporting, as well as delayed reporting, to not only better help survivors but improve education and policies surrounding reporting.
Of course, the “could” and “should” of reporting are two entirely different things, and reporting may not always be in the best interests of the survivor.
“The other really significant thing to keep in mind is that reporting to the police and going through the trial process can be highly traumatic for survivors – it is often referred to as the ‘second rape’,” says Fileborn.
“Given all of these social/cultural and institutional barriers, it is entirely understandable that the vast majority of survivors never report or delay reporting.”
Of course, this doesn’t dismiss the importance of identifying trends in under-reporting of sexual assault. Fileborn says the machine-learning approach can play an important part in identifying service support gaps in certain locations, and in increasing the availability of and access to these services.
“It’s possible that continued non-reporting might also tell us something about either the strength of community misconceptions about sexual violence and/or how police are responding to reports in particular locations.
“At the very least, this type of spatial analysis might help to flag locations that require further exploration.”
If you or someone you know is impacted by sexual assault, domestic or family violence, call 1800RESPECT on 1800 737 732 or visit 1800RESPECT.org.au.