Design Solutions to combat the spread of Misinformation and Fake News
Strategies to improve the quality of the content we engage with online
The value of validated information is critical infrastructure for any functioning society. Proper news outlets verify the sources of their information prior to publishing. This establishes trust in their content for readers — building reputations for publishing factual narratives, and earning the trust of readers as “objective” news sources. We learned to trust journalists, largely because they fact-check rumours. ‘Journalistic standards’ were a thing, blatantly false information was rarely proliferated at a large scale.
This has been undermined in recent years since the mass adoption of social media. Millions of people now find themselves in new, targetable, frictionless environments with an increased reach than ever before. Publishing content is now a democratized, zero-cost endeavour. Curation algorithms rate content based off engagement rather than reliability. This further incentivised sensationalist and controversial content to proliferate. Without the same ‘journalistic standards’ it has led to a rapid increase in impactful viral misinformation. It has become very difficult for people to distinguish fact from fiction with information they get online.
Whether for financial or political gain, fake news has now become more widespread. Research found fake news is 70% more likely to be shared. It was also found that false news spreaders tended to have fewer followers, followed fewer people, were less active, less often verified and were on the platform a shorter time relative to other users. Bots and recommendation algorithms have also been shown to increase the spread of fake news.
As elections, economy and public health narratives continue to be manipulated with misinformation, it’s clear that current answers aren’t working. Prevention is better than the cure. Retroactively combatting viral rumours through content moderation and takedowns are inadequate. New developments in platform design are necessary to create a more trustworthy and informational social platform. Enter Waivlength.
How Waivlength will combat misinformation/fake news
1. The Share Button
One-click sharing can allow content to spread at a very rapid rate. This often allows information to permeate very quickly before it has been verified. Sharing has become too easy and is often done quite emotionally, without real thought or consideration. Interestingly, around US election time when Twitter added some friction to discourage thoughtless sharing by prompting a commentary on shared material, it led to a 20% reduction in shares. Similarly, when WhatsApp put restrictions on forwarding of highly shared messages to a single reshare to stem the spread of COVID misinformation in April 2020, it reduced viral message forwarding by 70%.
One of Waivlength’s solutions to this is to limit Share options for posts. Sharing will be available via direct messaging more readily, but there will be scarcity to public sharing options. Public share options will be limited to 1 per day and require meaningful interaction akin to a Quote Retweet. This will promote more thoughtful responsible posting and promote the nudge proven to be effective for reducing the spread of misinformation online.
2. Labelling
Some fast-spreading content can be more dangerous than others. Entertaining and creative content can be celebrated with viral acknowledgement, however, when it comes to content related to politics, breaking news, healthcare etc. it is important that there are ‘viral circuit breakers’ in place for content to be appropriately fact checked and verified.
On Waivlength, content creators will choose from pre-set labels/tags to describe their content when posting. This can ensure information is interpreted in the right context and creates an environment where accuracy is rewarded. Opinion pieces can be differentiated from fact-checked articles.
Content creators can gain reputations and followings for producing reliable informational content, rather than be simply rewarded for driving engagement via likes, comments, and shares. Labelling can aid in maintaining the integrity of the site, provide better data to inform algorithms and machine learning, and create easier navigation to users towards content they seek.
3. Engagement Options
Part of the issue with increased sharing lies in a lack of interaction options — just because content feels more meaningful than a simple ‘like’ or ‘favourite’ can represent, it does not mean it warrants being shared with all followers/friends. Alternative options to interact and credit a post could help reduce the volume of shares it gets.
Waivlength will have more interaction options to credit and label content than a simple binary ‘like’. Users can rate content using a linear sliding scale to more accurately reflect their enjoyment/approval of the content. This makes the quality of content more apparent.
For low quality content that is repeatedly rated lower, the AI recommendation algorithms will ensure the post becomes less visible to other users.
4. Community-based Moderation
Harnessing the wisdom of crowds is a subject that’s been researched a lot and shows huge promise. Crowdsourcing from a group of 1,128 of users, researchers were able to segment groups as small as 10 individuals online that could accurately determine whether an article was false — about as well as professional fact-checkers. A recent study has also yielded positive results in using the crowd to verify or debunk claims far faster than professional fact checkers, with similar levels of accuracy.
The main factors which allow this to prosper is to preserve independence, diversity, and equality when processing crowd opinion. This is very difficult to do when rating is done in the context of a social network. However, crowds made interdependent by social influence can still exhibit wisdom — even greater wisdom — in the right environment.
Waivlength intends to harness the power and wisdom of its community as much as possible. A community moderation rewards pool of $WAIV will be available as an incentive for accurately flagging inappropriate content and moderating the site through its unique ‘jury-duty’ model. Much like Wikipedia, design solutions for harnessing the wisdom of crowds for a collective good is a priority.
5. Machine-Learning and Recommendation Algorithms
While most platforms rate content higher based off the shares and comments it receives, this often rewards the creation of controversial and emotive content. News articles that provoke surprise and anger are those that generally spread most quickly, thus rewarding provocative, controversial, and often misleading content.
On Waivlength, complimented by the labelling of content by creators and the community, algorithmic design can be such to reward content appropriately based off its individual category. For example, content purely designed for entertainment should be given the opportunity to go viral based off engagement and likes. However, news articles and opinion pieces should be ranked higher based off the quality of information and veracity of the content.
Human labels are essential for training machine-learning algorithms and to ensure that human judgement leads algorithmic judgement in defining truth and falsity.
Conclusion
The ever-increasing body of research on how social media has and continues to influence our society has taught us a lot about how we can harness its powers for a brighter social age. With continued technological advances in state-of-the-art machine-learning algorithms, we are now in a great position to learn from our previous mistakes and design a platform where a collaborative ecosystem reigns supreme. Decentralised platforms have the power to harness its community better than any.
Waivlength wants to promote a trustworthy and responsible online space. The design of the platform will ensure it facilitates this in many ways. Creating strong barriers against low-quality and/or false information spreading is a key component of this.
Forthcoming Updates
Waivlength is a grant recipient from the Algorand Foundation. While developers and advisors continue to work hard over the months ahead on platform build and securing external investment, it is clear that this platform has the potential to make a huge global impact as a competitor to current mainstream social media.
Learn more at www.waivlength.io where you can find a more detailed whitepaper and roadmap for the development of the platform, sign-up for the launch of the dApp and find contact details for the team.
Special Mention
Big thanks must go to Sinan Aral (Twitter — sinanaral) whose research and 2020 book The Hype Machine was a very informative source of information for this article. Also, a special word of thanks for Tobias Rose-Stockwell (Twitter — tobiasrose) and Renee DiResta (Twitter — noupside) whose content and ideas were also a great source of inspiration. Your continued work to improve the future of the Social Media landscape is admirable.
