App Freedom
A direct approach to a human-friendly internet
hawke.org/sandro
appfreedom.org/2023/dwebcamp
The internet is great
- Fun global culture
- Wikipedia, Crash Course, GitHub
- Disaster relief
- Exposing corruption
- Remote collaboration
- Economic growth
The internet is bad
- unfair bans
- empowering people doing harm
- surveillance
- corrupt algorithmic bias (tuned for outrage)
- dark patterns (unwanted features)
- unwanted ui changes
- unwanted policy changes
The underlying problem
- Each app has its own community
- Each app has its own database
- Each app has its own policies
- Each app has its own police
- Each app is a kingdom (with walls)
Social Computing = Vendor Lock-In
Considering moving away from Twitter?
- Lose your feed
- Lose your followers
- Lose your posting history
- Lose your conversations
- Lose your algorithmic training
- Lose your community
App Freedom
- Switch apps like switching salad dressings
- Try five new Twitter clones in a day
- Always the same users, tweets, replies, retweets, likes
- Different UI
- Different algo
- Different restrictions, features
Not Just Twitter
If it bugs you, and you haven't switched.
- YouTube
- eBay
- Uber
- RunKeeper
- Tinder
App Freedom
Users move seamlessly between competing apps
- Data under control of the user, not the app
- Relationships are part of the data
- Competing apps interoperate
- Communities cooperate for safety
- Competitive markets for apps, hosting
| problem | how it gets (mostly) solved |
|---|---|
| unfair bans | users pick their host |
| empowering harm | tools for community defense |
| surveillance | users pick host, diff incentives |
| corrupt algo | users pick their algos |
| dark patterns | users pick their apps |
| unwanted ui changes | users pick their apps |
| unwanted policy changes | users pick their apps & hosts |
AppFreedom.org
Let users move seamlessly between competing apps
Part 2
A Technical Solution
Four key technologies
Iterating on the problem
- Data stored under user control
- Data translators (instead of standards)
- Network of credibility assessments
(key to community defense) - User data sandbox
1. Data stored under user control
(diagram from 2012)
(diagram from 2012)
Blog-style decentralization
- Every data-write is to the user's web space
- Data reads/queries are across many users
- With host-to-host authn / authz
We're developing SiteQuery protocol
Other options available
2. Data translators
(instead of standards)
App schemas
- Every app stores its data using some schema
- To work together apps need compatible schemas
- Maybe market leader defines it (and controls market)
- Maybe open standards org (slow and expensive)
- Often apps just don't work together
Translation shims
- Each app uses whatever schema it wants
- When published, apps must include their schema
- When a user moves between apps, the system looks for trusted translators ("shims")
- From schema-of-writing
- To schema-of-reading
- Runs in a special sandbox
Ecosystem
- Trust network to find shims
- Private incentive to switch apps
- Commerical incentives to buld shims
- Possibly suited to LLM authoring
- In time, de facto standards may emerge
- But rapid innovation still welcome!
3.
Collaborative
credibility
assessment
Credibility assessment
- People enter some sources they trust (high cred)
- Maybe some they distrust (low cred)
- Optionally give reasons
- Optionally link to evidence
- Forms a recursive credibility network (trust graph)
- Your ratings let your apps know who to trust
- Argues to your community who they should trust
- Creates ecosystem with bias toward integrity
- Awaiting real user testing and experience
Might be the scalable and safe answer to online trust and safety
(Thanks to credweb.org participants, funders)
Also proves identity
- Build up your "crowd" out of testimony links
- Strong against impersonation
- Strong against bots
Does this increase division?
- Major concern, needs validation, but
- Disinformation increases division
- Makes it more clear you're in a bubble
- Nudges people to operate more rationally
- Brings evidence into the process
- Less driven by outrage-engagement
4. User data sandbox
- It's risky to give apps full read/write access to your data
- The trust graph helps - "only install if you trust..."
- Sandboxing is better
- Apps can read/query anything you can see
- Process and display to your screen
- They propose undoable updates to your data
- Surprisingly nice coding environment
- Performance seems good enough for many apps (so far)
Four technologies (summary)
- Data lives in user's webspace
- Data conversion shims for interop
- Credibility network for trust and safety
- User data sandbox allows untrusted apps
Next steps
- We're working on demo apps of whole stack
- Join mailing list / resource list
- Talk to me
- Why this won't work
- Good partnerships
- You can push for app freedom
AppFreedom.org
Let users move seamlessly between apps
Photo credits
Adrian Campfield
Nicholas Demetriades
Jan Blanicky
Victoria Pickering
Unknown PxFuel
Photos are linked to source page
App Freedom A direct approach to a human-friendly internet hawke.org/sandro appfreedom.org/2023/dwebcamp