Anti-Polarization Social Platform

A social media platform based on social science with blockchain incentives to build social discourse, and detect adversarial info operations

Invented first intelligent agent social simulation and award winning neural net agent simulations of polarizing Info Ops in social media

We will build a better Facebook that relies on beneficial algorithms instead of armies of people that need to prune bad content that is incentivized by bad profit seeking strategies.

Social media is a poor proxy for social interaction precisely because it is not social. Its algorithms are not informed by the protections of the natural social interaction it is replacing, nor of the nature of social institutions it is breaking. And this is because social media in a competitive market is not incentivized to consider these issues. Because social media has not preserved the protective natural social institutions that uphold social discourse, and effects on society were not measured against social science knowledge of the benefits of natural social interaction, social media spreads hate and divides populations. Polarization occurs even without adversarial information operations that intentionally break institutions for political gain, but information operations magnify this polarization greatly.  This ecosystem that is motivated and powered by profit-maximalization ensures that socially destructive algorithms will be hard to eradicate with regulation.

Fortunately, blockchain's ability to change the very infrastructure of trade offers hope. Blockchain's distributed ledger is a new infrastructure of commerce that can be used to modify incentives in a free and open market. In this project we first implement a market of modified incentives that promote a new business model in a social media platform, and next we detect the effect of our new social media algorithms on institutions so that natural social protections may be replaced and enhanced. Finally, we shore up our social media algorithms with ways to detect and protect against adversarial information operations, which they are vulnerable to even without perverse incentives.

One way to change social media's business model from making a profit by presenting products based on how much time they make available for ad viewing to one based on social good is to let consumers state their value preferences and enforce them in the display of ads. Like current social media companies, the platform will serve as a market that shows them goods that they like, however it will present ads for products to people based on how they fill their stated, measurable value preferences. One of these value preferences may be that it was not advertised in social media with divisive content, another may be that the products did not harm the environment, and another may be that the people who helped to create a product are all equitably compensated. Our AI can score products according to how well these values are kept using the distributed ledger. Only simple smart contracts are needed for this, and not even at the expense of value for price. For example, simple pareto optimization of social and economic value in a marketplace can promote much more social value than the present system where profits are required to be the sole consideration in business decisions.

The power of the market will convince smart contract designers to spend as much effort in ensuring that the unintended consequences and second order effects of releasing social media algorithms into society are made just as fail safe as the prevention of a double spend. We also offer a way to measure social consequences of algorithms with intelligent agent social simulations. These consequences include polarization. We have special complex adaptive system techniques of coevolution in which agents mimic societies based on incentives, and in which new incentives structures can be introduced to capture the effects on societies in silico. In particular, emergent social institutions from reinforcement learning agents with models of human cognition can help us analyze how smart contracts will affect human psychology before they go live.

These techniques will also be used to detect and protect our media against intentional polarization. Our coevolving agents have been applied to the problem of polarization in award winning studies at US Army and the US Office of the Secretary of Defense. In one model of tribalistic and divisive information operations on agents that each have a neural network mental model of cognitive dissonance, insurgents succeeded in gaining popularity by increasing divisions through disinformation campaigns on different factions of society. Applying these simulations to test social platform algorithms for their ability to magnify or to dampen information operation campaigns can measure both polarization and its opposite, social cohesion and the protection of social institutions. Only with such measurement can protection of social structure be incentivized through smart contracts that prioritize goods based on social value preferences.

**How do we address the challenge question?

What do SingularityNet's technologies make possible?

Singularity Net makes possible the combination of blockchain, a social incentive system to create new decentralized organizations of social relationships, with AI, to measure the success of these new social organizations. This proposal uses this principle to redesign the business model in social media ad markets through AI monitoring of blockchain's distributed ledger and AI monitoring of social media algorithms for social effects like social coherence and polarization.

How and where can AI + Cardano benefit society?

We can promote social good through infrastructure change, the market infrastructure and the capacity to foresee the consequences of the smart contract rules on social institutions before they are launched.

**Qualifications of Principal Investigator

I am the CTO of a Singularity Spinoff for social good, Rejuve.  I have a doctorate in Computational Science and 30 years' experience in AI for social science.  I wrote the first agent based social simulation in 1991, published in Behavioral Science and featured in The Economist. I have worked on Micro Macro Integration of human psychology and human institutions throughout these years, particularly in my SISTER simulation (Symbolic Interactionist Simulation of Trade and Emergent Roles) which I have rewritten at Singularity net as the Singularity Net Simulation and will use as the base simulation for this project. (See GitHub link above)

I have worked on designs for blockchain for social good at singularity net (see https://docs.google.com/document/d/1UAdANUE7mGjTeJtELboRcvDwOKNUXqNJbgQimUM2mXY/edit?usp=sharing and  https://docs.google.com/document/d/1dDZrnpHjzYXx6EsU39lawMqZUeppqt_lFmppzmJp0P4/edit?usp=sharing  ), and have made a NFT based tokenomics  design for equity at Rejuve (see  https://medium.com/rejuve-io/rejuve-the-decentralized-ai-powered-longevity-research-network-703e5631dcc1 )

I wrote award winning simulations of institutional coherence for the DoD, including information operations for polarization. My Nexus simulations of polarization (Nexus Schema Learner) and Corruption (Nexus Network Learner) is described in this Mitre book of the DoD Human Social Cultural Behavioral Program (HSCB):  https://www.mitre.org/sites/default/files/publications/sensemaking-ch13.pdf . The Nexus Models were used in US Office of Secretary of Defense Wargames and at the US Army Irregular Warfare Tactical Wargame, which won the US Army Payne Memorial Award for Excellence in Analysis in 2011, and was used in training for Information Operations: (see https://home.army.mil/wsmr/index.php/about/news-home1/research-and-analysis-center-celebrates-35th-anniversary https://www.army.mil/article/70181/new_war_game_developed_to_study_armys_impact))

The data absorption technique that I developed as part of this effort will be used to support the online simulation that monitors the social media for information operations. See ( https://computationalsocialscience.org/wp-content/uploads/2013/08/Duong2013.pdf )

This proposal simply pieces together work that I have done previously, and so can be done in efficient time.

For more information, see my LinkedIn page: https://www.linkedin.com/in/deborah-duong-02a0aa49/

**Breakdown of Budget Requirements

Tasks

Task 1. Design and implement Proof of Concept social media platform market ad ranking based on consumer social value preferences including polarization and social coherence

Task 2. Design and implement Proof of Concept Social Simulation with which to pre-test social media algorithms for effects on social institutions including polarization and social coherence

Task 3. Design and implement Proof of Concept online monitor of information operations utilizing simulation of task 2

Task 4. Final Report

**Expected Public Launch Date: Integration with social media platform by 12 months after award

**Budget

Task 1. One AI for Social Science expert 2 months, 10 hours per week at 90 dollars per hour 7800 USD

Task 1. One AI dev 2 months, 20 hours per week at 45 dollars per hour 7800 USD

Task 2. One AI for Social Science expert 2 months, 10 hours per week at 90 dollars per hour 7800 USD

Task 2. One AI dev 2 months, 20 hours per week at 45 dollars per hour 7800 USD

Task 3. One AI for Social Science expert 2 months, 10 hours per week at 90 dollars per hour 7800 USD

Task 3. One AI dev 2 months, 20 hours per week at 45 dollars per hour 7800 USD

Task 4. AI for social science expert 2200 USD

Total 49000 USD

**Definition of Success

3 months: Completed Social Media market and ad ranking system POC

6 months: Completed Simulation Analysis of Social Media Algorithms for polarization and social coherence POC

12 months: Integration in Singularity Net's Social Media platform