Focus on Features
A project of the Integrity InstitutePlatform design has the transformational capacity to shape (and reshape) the digital ecosystems in which we exist. A well chosen set of principles and approaches can allow platforms to simultaneously build user trust, and increase their public perception as a place where users' interests are respected and their safety is guaranteed. A poorly designed set of features has the potential to erode user loyalty and trust, and degrade the brand of an online space into irrelevance.
Intervention Design is the creative re-imagination of how platforms could operate in the future, and it's an imaginative exercise that puts legislators, regulators, and every-day users into the seat of product manager: how should the platform work, and what are the trade-offs implied by that design?
This exercise is far from hypothetical. Most of the interventions listed by this site could be demanded by users, enforced through legislation, or required by regulators.
Once we have the ability to imagine a better internet, we will find the power to build it.
This page features a long list of ideas of how platforms could change. Though most of these will probably sound too radical, too technical, or too invasive to seem realistic to you, I suspect a few will seem exiting, impactful, and possible - and if a better internet is just a few ideas away, what could you do to make it happen?
Omit comment reaction volume
Don't prominently display the number of likes or other forms of feedback a comment gets.
Comment Ordering
Set the tone for the conversation by choosing the order in which voices are heard.
Temporal Comment Limits
Restricting the volume of comments people can post will make folks think twice about their actions.
Affinity To Comment
Require a history of interaction between users before they're allowed to interact in comments.
Comment Tone Check Popup
Before posting content containing vulgarities, prompt a user to think twice.
API Surface Area
Prevent automated access to features that are intended only for human activity.
Auto-Generated Content Quiz
Encourage users to dig deeper than headlines by validating their knowledge of the subject.
No Comments for Fresh Accounts
Don't accept comments from newly created or barely used accounts.
Author Comment-Moderation
Enable primary content creators control over the comments layered on their content.
Hide Interaction Counts
Foster authentic interaction by making numerical properties less central.
Limited Number of Subscriptions
By limiting the number of subscribers (or the number of subscriptions), a platform can design toward real-world connections, and away from exponential scale distributions.
All Subscriptions Reciprocal
Require "following" to be bidirectional to avoid an exponential distribution of reach and attention.
Posting Limits
Limit the volume of information a user can generate in a day.
Uniform Reach
Vary posting limitations in line with subscriber volume to achieve a uniform capacity to generate impressions
User-Defined Subscription Sorting
Enable users multiple different options for how to prioritize and filter content from their subscriptions.
Self-imposed time limits
Allow users to set hard cutoffs on platform use.
Mix in Authoritative Content
Add authoritative content from trusted partners on users' subscribed topics.
Request to Follow
Make Following a two-sided operation, so that celebs have implicit limits on the ability of fans to follow them.
Limit Time on Platform
To minimize the potential harms of a platform, enforce global time limits on engagement.
Configurable Feed Optimization Goals
Allow users to choose what they want their feeds to optimize for.
Perform basic link vetting
Run basic validation on the contents that something links to before showing the link to the user.
Warn Before Risky Action
Use signals about affinity and content to occasionally warn the user about what they're about to see/download/visit.
No E2E Encryption
When content is end-to-end encrypted, platforms can offer no content-based protections.
Two-Factor Authentication
Authenticating users through two types of credentials (something you have, something you know, something you are).
Identity Verification
Require users to register for an application with a state issued identity document.
Must Request to Message
Only allow friction-less initiation of a conversation between established connections.
Third repost has to manually enter the content
Slow the spread of virality by adding friction on repost/re-share functionality.
Recommend Only Verified Users
Require identity verification before adding content from an account to a recommendation engine.
Delete Abandoned Accounts
Prevent high-value usernames from being squatted on by periodically cleaning up inactive accounts.
Cross platform username verification
To create an account with a popular username, require proof of ownership on another platform.
Limit account volume
Reducing the volume of accounts a person can create restricts their capacity to cause harm at scale.
Right-size content visibility
Place limits on the amount of harm content can cause on a platform by restricting its reach.
Don't support large files
Large, high quality, high volume uploads are both risky and expensive.
Subtly Modulate Uploads
Features that provide exact replicas of the data in are ripe for abuse.
Limit Content Reach
Put upper bounds on how many people can view/share/interact with content.
Don't collect location data
Limit storage, retention, and sharing of location information including IP address.
Don't allow posting of Location
Use content filters to prevent users from posting addresses, latitude/longitude coordinates, or other location data.
No Search by Location
When abusers can discover content and accounts by location, the creators of content are in danger.
Allow immediate location takedown
Enable any user to immediately censor a physical address from public content.
Adults Can't Message Kids
Facilitate a sense of safety for parents + kids by only allowing prior connections to initiate conversations with them.
Require unleaked passwords
During signup, don't allow users to use a password that has been included in a leaked dataset.
Use an OAuth Provider
Centralize identity management + risk with a company that thinks about it full time.
Feedback Only From Authoritative Sources
Allow users with geographic proximity, purchase history, or other signals of engagement to leave ratings.
Two-Factor Payment Identification
Require one user to specify multiple pieces of information about another user to send them a payment.
User Submitted Reviews
Platforms that enable repeat engagement can benefit from allowing users to learn about the past experiences of other users.
Label/Detect Identical Content
For some features, duplicate data suggests misuse.
Reporting Mechanisms
Allow users to flag content or behavior that they find to be abusive.
Three Insult Rule
Rather than looking at whether individual pieces of content constitute harassment, consider patterns of behavior.
Flatten Virality Curves
Cap the attention a user can receive by a multiple of their prior reach.
Periodic Preference Reset
Regularly reset the profile used to generate recommendations for a user.
Ban Proactive Content Recommendation
Prohibit infinite feeds for children, and provide a universal opt-out for adults.
Read Before Re-share
Prevent gutteral reshares by adding a waiting period estimated by the length of content.
Require Physical Validation
Before leaving a review, require a photo of the reviewed entity.
Weight Reviews
Use the quality of previous recommendations to weight feedback across users.
Read before Comment
Require users to read the primary content before they leave comments.
Media Provenance
Record and display the chain of custody and original source for media.
Crowdsourced Annotations
Allow users to add context to the posts of others when it is widely seen to be useful.
Crowdsourced Metadata
Allow users limited access to content metadata, and enable them to highlight this on the posts of other users.
Bulk Location Takedown Tools
Allow a user to scrub their profile of location data while leaving other data intact.
Three Strikes: Warn, Suspend, Ban (USA)
When users report unwanted sexual advances, use a three strike system.
Save Extensive Metadata
Recording extensive metadata about sensitive activity can reduce criminal appeal.
Deprioritize User Identity
In platforms where the identity of the participants isn't central, omit it.
Anonymous Limitations
Require users to create an account before they can use features that create data or interact with others.
Require Labels on AI Created Content
Enact legislation for the mandatory prominent disclosure of AI generation.
Loading...