This article was originally published on April 18, 2016 in The Guardian. It is re-published here in accordance with a joint publishing agreement.

Jude Milhon, a cyberfeminist who died in 2003, was one of the first women to witness online harassment. Writing in Wired magazine, she urged women to “toughen up”. “Whether we’re set upon by zealots or bigots or abusively correct politicos, we have to learn to defend ourselves,” she said. The year was 1995, and Jude already had 20 years of experience with harassment. In the early 1970s, she was active on Community Memory, a digital classifieds service in libraries and record shops in Berkeley, California. To reduce the abusive comments, the system charged 25 cents per post.

Online harassment might feel like a recent issue, but it’s an enduring problem with 45 years of history. In 1984, 13 years after the invention of email, the social psychologist Sara Kiesler found computer manuals lamenting people’s terrible behaviour online. In the 1980s subscribers on The WELL, an early internet forum, developed block lists to ignore abusive messages.

The issue became a widespread public concern in the 1990s. Julian Dibbell’s article A Rape in Cyberspace caught the eye of the law professor Lawrence Lessig, who then wrote a book about internet governance. The writers of Wired Women published first-person essays on women’s experiences, including harassment and peer support. By the end of the dotcom era, companies relied on tens of thousands of moderators to maintain online relations, a mix of volunteer and paid work that continues today.

In the 2000s, web companies repositioned themselves as “platforms”, in part to limit their liability for users’ illegal activity. As more people came online, social platforms scaled human relations, leading social worlds to collide in ways that the sociologist danah boyd described as “context collapse”. Controversies were often amplified by metrics-driven publishers struggling for advertising revenue, according to the cultural studies professor Whitney Phillips in a recent book on trolls. Complicating our understanding of antagonism online, networked social movements now routinely leverage public emotions for political ends.

Harassment is not a disembodied, internet-specific risk. Among the 40% of internet-using US adults who have experienced it, half report that an unknown person was behind their last experience of name-calling, embarrassment, stalking, sexual harassment or personal threat. The other half, perhaps as many as 41 million people, have been harassed by acquaintances, friends, family members, and former romantic partners, according to Pew research in 2014 by Maeve Duggan.

The issue is rarely out of the news. High profile women speak publicly about the harassment they face. News publishers sometimes close down comments. Advocacy groups draw attention to online threats, racism, and discrimination. Platforms policies are frequently under fire. Commentators warn about the risks of shaming and censorship.

If you sense growing frustration and despair, it’s not just you. The incidence rate of weekly mentions of online harassment in the UK media has increased by 3.9% per month on average since January 2014 (data and analysis here).

“Research can shine light on where the best place is to put your fear,” says the former Pew researcher Amanda Lenhart. Now at the Data and Society Research Institute, Lenhart has spent 17 years surveying Americans about their online experiences. “Media coverage focuses on things that people pay attention to, often things that are frightening,” she points out. As panics about online risks come and go, researchers study an issue “in the data and not just in the stories that we hear”.

Online harassment is not isolated to women or feminists, as decades of research and experience have shown. It’s a pervasive issue for millions of people, a complex web of problems with no easy solutions. By now we know that avoiding the internet isn’t possible, and that ignoring social problems won’t make them go away. What does that leave us with?

In search of inspiration, I followed the history of public safety from another period of rapid change in a life-sustaining network, the transition to industrial food production in the Victorian era. The 80-year history of efforts to study and respond to food contamination holds many lessons for our response to online harassment.

Every meal we share connects us to a social web of food, according to the archaeologist Martin Jones. Each table is a “depository of information and engine of its transfer … spread across space and time”. Even the simple ingredients in bread connect us through history, evolution, and the global supply chains that were just reaching the working classes during the Industrial Revolution.

By the mid 1800s, British food production was moving out of villages and into urban manufacturers, where intense competition put pressure on food producers to reduce costs by adding cheap adulterants. By 1820, the chemist Frederick Accum documented what he called “culinary poisons” in London’s food. Lead additives strengthened the colour of chocolate and cheese. Tea and coffee were adulterated with sand, chicory, and spent grounds treated in ferrous sulphate and sheep’s dung. Not all adulterants were poisonous but, according to the food historian Annie Gray, they all contributed to widespread malnutrition in industrial towns.

As supply chains become longer, they also became more complex, open to adulteration at more than one point. Bakers would add chalk and alum to flour, unaware that the millers had already done so. In Christina Rossetti’s poem Goblin Market, unsuspecting women are lured to eat deadly, adulterated fruit from unknown places:

We must not look at goblin men,

We must not buy their fruits:

Who knows upon what soil they fed

Their hungry thirsty roots?

In Rossetti’s poem, fears of food adulteration merge with wider panics about sexual assault in Victorian women’s lives. As the structures of industrial life moved beyond local, face-to-face relationships, people lost faith in their ability to judge purity from appearance or to uphold community norms when problems were detected. Nor could governments help. In the 40 years between the first adulteration research and Goblin Market, food quality was mostly unregulated. With no accepted methods to identify adulteration, the character and scale of problems were still debated. They might just as easily have been discussing online harassment.

Today when industrial-scale food adulteration occurs, such as the 85,000 tonnes of olives treated with copper sulphate that Italian police seized in February, we expect a systemic response. Creating systems of public safety that also worked for industry was a complex endeavour that unfolded over generations, through the collective efforts of scientists, advocates, industry groups and governments.

A workable approach to food safety requires methods of detecting unsafe food, methods that still weren’t obvious in the 1850s, 30 years after Accum’s first report. That changed when several London doctors joined forces to create the Analytical Sanitary Commission.

By 1854, the London doctor Arthur Hill Hassall had collected 2,500 samples from across the city, analysing them with the medical chemist Henry Letheby using novel microscope and chemical tests. The name of each shop was recorded, and proprietors were warned that they would be publicly shamed if they didn’t change. When Hassall’s reports were published in the Lancet, they showed that “adulteration was the rule rather than the exception”. Like Pew’s study of harassment, the Analytical Sanitary Commission had used research to identify a cause for public concern.

Reliable detection of adulteration invigorated campaigns for food safety, but a systemic response took several more decades to establish. Some argued for government regulation, but others encouraged communities to self-regulate, proposing a microscope in every home. Parliament chose a middle way, providing funds for regional food safety analysts in 1872, almost 20 years after the Lancet’s first reports. Only then could professional societies such as the Association of Public Analysts train experts for these new roles.

In the US, food purity regulation took even longer, led by 19th-century feminist crusaders. As women encountered opposition on food and drug safety, they learned to organise politically to reform local ordinances, science education, and national policy. In 1876, the MIT instructor Ellen Swallow Richards, the founder of home economics and a pioneer of public health research, started a women’s lab to train female chemists and citizen scientists. The women’s magazine Good Housekeeping also built a lab much like the Lancet’s Analytical Sanitary Commission. The US finally passed a national law on food and drug safety in 1906, 30 years after the UK.

Companies and industry associations also hired their own researchers, competing with advocates and other companies over which products were considered safe. In Ireland, one industry researcher made a fundamental breakthrough in science. The Guinness Research Laboratory hired William Sealy Gosset in 1900 to test ingredients and brewing processes. Gosset needed a mathematical way to test claims about quality without checking each kernel of wheat or barrel of beer. Guinness gave him intellectual independence to publish any scientific advances under a pseudonym and collaborate freely with academics.

Gosset’s invention, the t-test, gave Guinness the lead in scientific food production and changed statistics forever. It remains a basic method in medical trials, economics and survey research today.

Gosset used the t-test to define margins of error for acceptable saccharin levels in malt extract, which influence beer’s alcohol content. A hundred years later, when researchers such as Lenhart study online behaviour, the t-test offers guidance on how many people to survey for representative knowledge about the entire population.

When Gosset published his method in 1904, the struggle for food safety was far from over. But 84 years after Accum’s early findings on adulteration, the UK and US were arriving at institutional, legal and scientific arrangements that worked for citizens and businesses alike. When new health controversies arose, an ecosystem of citizen groups, regulators and industry experts could assess the situation and work together to keep the food supply safe.

Today, our efforts to understand and govern social problems online are at a similar point to 1850s food safety campaigns. Based on decades of research, we have a few early-stage methods to study these problems, and we have fragments of conflicting knowledge about effective responses. A growing number of experienced advocates, researchers and platform safety teams are paying attention. We still lack a working ecosystem of arrangements to handle emerging problems and maintain the public trust though.

In online harassment, the power of single stories can lead us to misidentify a problem. “It’s often white women who are seen as the targets or somehow the victims that we are invited to be concerned about,” the City University of New York professor and cyber-racism expert Jesse Daniels points out. “When we focus on white women, race drops out of the conversation.” Online harassment isn’t limited to women. Pew’s 2014 study found that while greater percentages of US women are sexually harassed and stalked online, men report receiving physical threats at higher rates than women. People with LGBTQ identities are only occasionally consulted in systematic internet research.

Data on social problems is only as good as the questions we ask. Computer scientists often assume that prejudice and hate are deviant behaviour outside mainstream society. Findings from qualitative research dispute that assumption. “Racism is not an outlier, a deviant perspective. It’s core to American culture and values,” says Daniels. Writing about networked harassment online, the University of Southern California researchers Sarah Banet-Weiser and Kate Miltner argue that “misogyny is not only widespread and deeply entrenched in western culture, it is naturalised”. The underlying causes of online harassment can’t be solved by detecting and banning a few toxic commenters.

What we learn from the fight against food adulteration is that persistent, concrete steps in research, policy, and social action can grow into systemic change. Enduring progress requires fundamental social change, even as we pursue immediate protections through design, law and community norms.

Professor Cliff Lampe, a computer scientist at the University of Michigan, published some of the earliest studies about content moderation online. In the early 2000s, “Slashdot was one of the first sites where you saw big management problems,” he says. The site, which was 94% male, was threatened by “a strong development of trolls and harassment”.

To see if readers could moderate themselves, Lampe analysed millions of Slashdot user actions and developed a systematic understanding of volunteer moderation. First, he studied how much moderation work was needed to detect comment quality and whether rating moderators could make the process fairer. Then he asked if moderators influenced behaviour, discovering that downvotes might help users learn. Lampe also studied how readers used collaborative filtering systems akin to block lists on Twitter. Overall, his findings were optimistic about volunteer moderation.

At a recent MIT workshop on comment systems, Lampe wore a rueful grin as speakers speculated on ideas that he had tested 10 years earlier. His experience, however, is not unusual among academics. Copyright restrictions, trade secrets and legal concerns often hold back public knowledge about online safety.

Most research about online harassment is inaccessible to the public for copyright reasons, including all of Lampe’s research on moderation. “Hiding our work behind paywalls” prevents the benefits of research from reaching the public, argues Daniels. When I recently led a team of scholars to review over 1,000 scholarly articles related to online harassment, very few were legally available beyond university networks.

Companies may also be holding important knowledge back. Market researchers and platform employees sometimes tell Lenhart privately that they have findings on her research questions but aren’t allowed to share them. Rather than advancing knowledge on public wellbeing, companies worry about sharing unflattering results. While many platforms share detailed data with researchers about copyright complaints, only Wikipedia shares survey information on the kinds of harassment its users report. This week, the Guardian became the first platform to publish behavioural data about abuse in Guardian comments.

Companies sometimes prevent academics from publishing at all. When studying comments on YouTube, Lampe signed non-disclosure agreements giving Google the right to “squash any paper for technical, legal or PR purposes”. He says his research “never got past their corporate lawyers”. Like many academics who collaborate with corporations, Lampe had made a bet. A relationship with YouTube might allow him to influence the platform for the better, even if his findings remained a trade secret.

A spokesperson for YouTube said: “While we can’t comment on this specific research, we consult extensively with NGOs and individuals who work on these issues so we can hone our understanding and develop policies that strike the right balance. In addition to our policies, we’ve developed moderation tools that allow creators to better control comments on their videos and we’re continually exploring new, innovative ways to address harassment on the platform.”

It’s tragic that secret research with for-profit corporations is often the only public service available to academics. This is not confined to YouTube, but is common industry practice.

Companies are responding to external forces when they avoid creating public knowledge on harassment. Public fears about privacy, censorship and the ethics of social experiments make online safety research a volatile PR risk. Lawyers sometimes fear that publishing on problems opens their companies to new liabilities. In consequence, good work inside companies goes unacknowledged and tested ideas can’t spread.

When research stays a trade secret, public safety suffers. Not a single field study on moderation or online harassment has ever been replicated in 40 years of research. It’s hard to know if Lampe’s moderation research applies anywhere other than early 2000s Slashdot. It took 10 years for others to ask similar questions, and their findings seem to point in an opposite direction. Which should we trust? With so little peer reviewed research and so few platforms willing to work in the open, basic questions about online safety have no conclusive public answer.

In the 19th century, food safety advocates circumvented companies to research health risks on their own. Online support groups and community moderators are doing the same. The peer support platform HeartMob is working with researchers to study the needs of support-seekers. Last year, the advocacy group Women, Action and the Media collected three weeks of Twitter abuse reports while supporting people who received harassment. On reddit, volunteer moderators of communities with up to 10 million subscribers are producing detailed transparency reports.

As media controversies about online harassment flare and fall, the careful work of research continues. Lenhart recently received a grant to study links between cyberstalking and domestic violence. “My approach has always been to be part of the conversation by injecting data into conversations where no data is currently being exchanged,” she says.

In Goblin Market, Christina Rossetti imagines a future where the experience of adulterated food would be a story that parents tell their children of “days long gone”, a morality tale about the courage of those who endeavoured to “win the fiery antidote”. Online harassment doesn’t have a simple antidote, but neither did food safety. Steady steps in research, advocacy, and institutional cooperation can move us toward a future where we don’t have to look twice before tasting the fruits of our increasingly connected world.