The first time I wrote an article for a newspaper, the first online comment said: 'If I ever see you in the street, I hope you get shot.' The article was about being abused and harassed in the street, specifically while cycling. I wasn't surprised that the online comments mirrored the behaviour the article addressed. But unlike the men who shouted at me as I waited on my bike in Clapham, the online commenter could be sure I wouldn't spit in his face in response.

I ended up moderating comments at the Guardian for two years: without a doubt, the worst job of my life. The time I spent washing heavily soiled laundry in a care home didn't affect my sleep; the insults hurled at me in a major bank's call centre were far milder. The comments that appear under most Guardian articles are barely worth your time; the ones that don't make it through are completely toxic. Hundreds of people, day after day, create multiple accounts to spew the most extreme racist, sexist and homophobic abuse imaginable.

At the same time, the volume of comments was increasing rapidly, but the number of moderators was not. The process became more automated and the focus shifted from 'steering a conversation' to simple damage control. Writers were encouraged to go 'below the line' and reply to comments, but the presence of authors, especially when they were women, encouraged commenters to act up and hurl more insults. The vast majority of commenters are men, and the Guardian’s own research shows that the writers most often abused on their site are women and black journalists, and the least abused contributors – surprise – are white men. (This is also true of the LRB blog.)

The patterns are replicated on a much larger scale on social media. Demos released a report this week on the extent of online misogynist abuse. Most women don't need a report to tell them how much a lot of men hate them: going out in the street tells us that. We know that smiling or pretending to ignore unsolicited attention from strangers is often the safest response; we don't answer back for fear of violence.

The Demos research analysed comments over a three-week period, using an algorithm to determine whether tweets that used the words 'whore' and 'slut' were sexist insults or 'conversational'. They found that 6500 users were targeted by 10,000 'explicitly aggressive and misogynistic tweets'.

Has the internet made people more hateful? Perhaps. Or it may simply have made it easier for people to express their hate. The first time someone called me a 'whore' I was 12 years old, walking home in my school uniform. He had to wind his window down and wait for me to pass his car. When I first got a mobile phone, I'd sometimes get abusive phone calls late at night. It all took the men responsible a lot more time and effort than sending a tweet does. Now, quite often men will tweet photos of their erect cocks to me, in response to nothing at all, and their profiles show they do this to as many women as possible, several a minute, before their accounts get shut down. It must be an easy way to get a kick without leaving the house.

People who don't get harassed in the street, or who have never noticed random racial or sexual abuse hurled at someone else, may think it doesn't happen very often. But you're more likely to experience racial abuse or sexual harassment and homophobic abuse when you're alone, and an easier target, so there are fewer witnesses. But not always: and when you are attacked in a place with witnesses, the familiar casting of eyes downwards and elective deafness tells you that, even though you're physically surrounded, psychologically and politically you're on your own. Most women and gay men I know won't get night buses alone now, because they've experienced this public abandonment and tacit condoning of abuse too many times.

But the ease of sending messages to strangers through social media has also flattened power structures for the better. When I worked for an MP, she got letters and occasional emails. It took people effort and thought to send them, and researchers and staff acted as gatekeepers. Most MPs now have Twitter accounts, which make it easier for them to publicise what they're doing, but also easier for constituents to get in touch and express their displeasure at their elected representatives' voting patterns.

Last December, for example, a lot of people got the impression that MPs on the right of the Labour Party voted in favour of airstrikes against Syria in order to undermine Corbyn's leadership, rather than out of a deep and abiding belief that it was the best way to save civilian lives and bring peace to the region. The deluge of complaint that followed surprised some MPs – and, predictably, many tweets directed at women expressed their anger in misogynist terms. Hate speech sticks in your mind, and has no place in public debate; Facebook and Twitter do too little to protect women and stamp out racism on their platforms. But you can't use the hate speech as an excuse to dodge the valid criticism.