Twitter has pledged to curb harassment before. In February CEO Dick Costolo admitted, “We suck at dealing with abuse and trolls on the platform and we’ve sucked at it for years.” Last week Vijaya Gadde, Twitter’s general counsel, promised the company would do better. It has hired more people. It has invested in new tools. It is overhauling its safety policies. Twitter, she insists, is going to change.

But it won’t.

Abuse on Twitter is a serious problem, and it is going to remain a serious problem no matter how Twitter tweaks its terms of service and no matter how many staff it has patrolling for violations of those terms. That’s because the bullying and harassment on Twitter is not incidental or accidental. It’s built into the structure of the service.

Twitter is distinguished from its bigger rival, Facebook, by the fact that it assumes consent. On Facebook, to talk to other users or see their posts, you generally first have to obtain their permission. You ask if you can be friends, and the person you ask either agrees or doesn’t.

Twitter is built on different principles. On Twitter you can follow someone unless they specifically decide to block you. Other people can follow you unless you take the time to prevent them from doing so. Most importantly, unless you specifically say they are not allowed to, anyone can talk to you and force you to pay attention. All they do is put your Twitter handle in their tweet, and you will get a notification.

The assumption on Twitter is that consent is unnecessary. You can talk to anyone — from Rihanna to David Frum to your cousin — without permission or barrier. This is a big part of the platform’s appeal. Facebook allows each user to create an exclusive club; Twitter creates a public square in which you can interact with anyone who chooses to set up a booth. That’s great for networking, great for conversation and great for activist projects, which can gather like-minded folks quickly and with few barriers.

But the same thing that makes the Twitter experience unique also makes it uniquely abusive. Twitter is essentially optimized for bullying, because, again, it assumes consent. When a celebrity dies, there are going to be a certain number of jerks who think it’s funny to taunt their family members — and every one of those people can do just that, directly, on Twitter. If you write about being a sexual assault survivor on a mainstream site, Twitter is where people can go to find you and shout sexual slurs at you, and all you can do is read each tweet and then block each user one by one by one. If a person with a large Twitter following decides they want to screw with you, they can just tweet your handle to followers and suddenly your mentions will be filled with abusive indignation. It’s not an accident that the floating harassment campaign known as #gamergate has centered on Twitter rather than on Facebook or some other platform. Twitter, by its structure, gives default power to bullies. It assumes people should be able to yell at you.

If Twitter wants to reduce abuse, therefore, it needs to take steps to be more like Facebook. At the moment, users must actively block accounts they don’t want to see, or they must actively lock their accounts to prevent people from seeing their tweets or following them or they have to use third-party tools such as the ggautoblocker, which uses algorithms to filter out most gamergate supporters. All of these methods are time-consuming and/or restrict access to the platform. It puts burdens on victims rather than on the harassers.

Twitter needs to empower good actors, rather than bullies. There are various ways this could happen. A simple start would be to only give users instant notifications from those they follow. All other notifications would go to an “other notification” folder — much like the “other” folder on Facebook. A more radical change would be to require consent for any use of a user’s twitter handle. If a user doesn’t approve the use of their handle after they’re mentioned, the tweet would not appear. This would undermine the performative aspect of Twitter abuse; there’s no point in yelling at someone for your friend’s benefit if your friends will never see it.

Some might object that alterations like this will change the character of Twitter. That’s true — and it’s a good thing. The character of Twitter has to change. Not entirely, of course; no one wants to erase the public square character of Twitter altogether. But Twitter does need to take steps to ensure that bullying is no longer a default assumption and that abuse from strangers is no longer assumed.

Instead, though, Twitter seems to be going backwards. It just announced a feature that allows people who don’t follow you to Direct Message you. In short, you can now get rape and death threats in your DM box as well as in your mentions. This is a horrible idea, and pushback seems to have led Twitter to reconsider. But even if the idea itself is rotten, it at least suggests that Twitter could, in fact, improve if it wanted to. The structure that enables abuse is a choice. If Twitter can tweak the rules to make harassment worse, it can and should tweak them to prevent it.

Noah Berlatsky edits the comics and culture site the Hooded Utilitarian and is a contributing writer for The Atlantic.