I am building an app to fight propaganda online. Need your help

You have the wrong idea about propaganda: most people don't use propaganda to make decisions, they use it as talking points to signal what kind of person they are. Politics are an aesthetic that most people wear, because at least in US, most people have the luxury to be immune for the most part from the federal government affecting them in tangible ways (this is also why this election was decided by 200 out of 350 million people that didn't vote).

So when you hear hardcore republicans talk about all the bad things that democrats are doing, its not that they believe that democrats are bad because they are doing those things, its that they fundamentally believe that democrats are intrinsically bad, and are using the talking points just to signal what kind of person they are.

This is why when republicans do the exact same stuff its either "fake news" or "different".

The only way to fight such ignorance, as ridiculous as it sounds, is ironically the opposite. You want to accelerate "propaganda" to the point where Republicans degrade the way of life in US to a point where people can't lead their lives day to day anymore without worrying about things like social unrest, lack of food, war, e.t.c.

My first attempt: "birds are real". Verdict: "false". Reading the generated blurb though made clear that birds are in fact real.

I love the idea and want to see this succeed. It may need a bit more time in the oven though.

Yeah I assume it thought that you are asking it whether birds are fake. In generated article it was always talking about claim that they are fake. Anyways, thanks for testing!

I was thinking about this and how we need a new generation of spam filters or content firewall to read the content and decide whether I would want to read it. Some sort of browser plugin

I also think we need policies around botfarms. I see all this free speech absolutism at the moment and think ok that's fine, but free speech shouldn't apply to bots or ai. At least at the moment

I tried a few more and it got them right. Good job! One nitpick is that the progress bar stays stuck at 100% for a few seconds before the results are actually shown.

clickable: https://truthorfake.com/

Seeme to do a good writeup and links to useful articles. You might want to have a feedback option for submitters: seems reasonable, versus analysis seems wrong bc details.

I heavily recommend to listen to the Alt Right Playbook series from Innuendo Studios. [1]

You have the right intent, but I think you're missing out on the game strategies at play. Personally, I decided to combat this war of misinformation in cyberspace rather than in mindspace.

The problem with their tactics is that they control the conversation, and that's the point. By controlling the narrative they control what's accepted (or pushed into being accepted) to talk about. Using bogus papers as proxies for their real intent, and themselves actually not being interessted in a scientific or rational discussion.

It's a war about beliefs and emotions, not opinions and discussions.

[1] https://www.youtube.com/playlist?list=PLJA_jUddXvY7v0VkYRbAN...

Sorry my ai credits got exhausting. Should be working back now

are these Russian bots in the room with us right now

first you have to define what is propaganda.

>>and it goes on web collects articles on topic then processes them with ai and makes a decision based on those articles.

Over 50% of which might also be misleading or blatant lies. Where would you find neutral sources?

Honestly, it seems to fail a lot on topics where there is a strong bias in the legacy media. There is also no "deduplication" of similar topics and related events.

I'd suggest a very different approach, expanding on the methodology of sites like "Here is the evidence" and "Ground Truth":

1. User asks question

2. If there are multiple possible matches, show them to the user and ask him to pick one, or, if not yet available or the matches are insufficient, ask the user to create a new request for clarification (including gathering some of the information he might already know, like involved people, approximate location, approximate time)

3. Once a specific event has been identified, show the user a summary of the event, a likely evaluation of his statement/question, a summary of all distinct positions on this topic and a list of the evidence gathered "here is the evidence"-style (but archived links..., and including a confidence rating)

4. If no good information is available yet, give the user an option to be notified via email or app or whatever when that changes

.

Upon a new request for clarification of an event being made (and this request not being merged with an already eisting topic):

1. Create a new entry, which does not yet make any judgement

2. Start looking up the event online, gathering neutral, relevant and easily available information like: Publicly known people present, relevant public locations (e.g. hospital victims were brought to in case of accident, location of event, involved agencies, nearby points of interest, nearby events), time, weather (helicopter crash during fog? Protest has low attendence during cold temperatures?), timeline leading up to the event (e.g. Presidential schedule, event happening after people leaving major sports event?, previous treaties), relevant laws and regulations, ..., basically everything major reports about the event think of as relevant and some basic other information (e.g. like the weather)

3. Update the entry with the information and ask people seeing it to gather additional evidence, maybe even provide commentary if applicable.

4. Notify relevant volunteers to gather additional evidence (ranked by living in the area, speaking the language, have specified this topic as an interest of theirs, have claimed that they have relevant abilities (e.g. access to sattelite images, participate in certain networks, ...), some randomly selected, ...). Maybe to avoid spamming anyone, default to sending a weekly list.

5. Depending on ressources and interest, keep gathering evidence on your own periodically

6. Gather evidence for the evidence. Plausible? Additional angles? Picked up by officials or involved people? People contesting it?

7. Give people the tools to contest, rate or ammend evidence. Through 6 and 7 eventually build a whole tree / graph of evidence and commentary.

8. Analyze the evidence where possible and relevant. Rate how significant it is, how controversial, how relevant (to legal proceedings, other evidence, logical conclusions) and how accurate (includes a spin? outright faked?)

9. Reduce the evidence trees down to a summary of the primary evidence. Summarize all the provided commentary, articles, etc. for each evidence into multiple distinct opinions.

10. Summarize all the primary evidence summaries and summarize all the evidence opinions and commentary on the main topic into distinct view points on this topic.

11. Update the entry with this information. Make a judgement.

12. Notify all those interested in the topic.

13. Periodically reevaluate if new updates / evidence becomes available. Periodically update those interested on major changes (or even minor if requested by those who are interested)