DeESG

Our project aims to decentralise the ESG process by leveraging AI agents, oracles, IoT data-collecting devices and Fully Homomorphic Encryption to ensure confidentiality.

The first problem statement includes bias in ESG evaluation. In this case, the current ESG scores can be manipulated by companies to present a better image, distorting the true impact of their activities. Secondly, another problem is the lack of trustworthy data. Current ESG evaluations rely heavily on self-reported data, which is often incomplete or inaccurate. There's a lack of verifiable, objective data to train AI models. Thirdly, lack of incentives. Companies are not sufficiently incentivized to improve their ESG practices, slowing down the shift toward more sustainable and ethical operations.

To tackle these problems, we will introduce our project called DeESG which will use unsupervised AI clustering to evaluate companies' ESG risk levels based on a broad range of environmental, social, and governance factors. This removes the subjectivity and potential biases that come with human-based evaluations. In terms of Environmental Data, IoT sensors will be utilised to collect reliable, real-time environmental data (e.g., carbon emissions, energy consumption) and send it to the blockchain. This ensures transparency, accuracy, and tamper-proof data. In terms of the Social and Governance part, employees can participate in a secure forum to discuss and vote on governance and social aspects of their company. Data from the forum, combined with FHE (Fully Homomorphic Encryption), ensures that employee feedback remains anonymous while still contributing to the company’s score. Another point which is worth highlighting is the Green Token (GTK). It is a tokenomics system that rewards companies for improving their ESG practices. Companies that demonstrate positive ESG behaviour receive GTK tokens, which are stored and claimed via blockchain smart contracts. The tokens incentivise companies to adopt more sustainable and ethical practices.

To explain how it works, it firstly involves data collection via IoT. IoT sensors (e.g., on a Raspberry Pi) collect environmental data at regular intervals (e.g., carbon emissions, energy consumption). This data is sent to the blockchain and verified. After verification, GTK tokens are rewarded to the device (smart contract) based on the data's authenticity. Only verified users can claim tokens.

Moving on, in terms of Social and Governance Feedback via Forum, every employee will have a unique NFTs as proof of employment. They participate in a forum to provide feedback on social and governance practices, such as work conditions, leadership, and corporate policies. Feedback is weighted by upvotes/downvotes from other members, and the AI evaluates sentiment. To do this, Inco Protocol's Fully Homomorphic Encryption (FHE) is used to ensure privacy while maintaining the integrity of employee contributions.

Furthermore, it involves AI clustering for ESG assessment. The AI uses unsupervised machine learning to categorise companies into different ESG grades. The model uses data collected from IoT sensors, forum sentiment, and other company metrics to assess how well a company performs in ESG areas. Over time, the AI model becomes more accurate as more data is collected, leading to better evaluations.

All data, including IoT sensor readings and ESG ratings, are stored on the blockchain for transparency and security. Smart contracts ensure that the distribution of GTK tokens is automated and tamper-proof.

The technology stack used are also of utmost significance to be mentioned. Chainlink is used for automating IoT data collection and interacting with smart contracts. The Graph is used to create decentralised subgraphs to retrieve on-chain data for the AI model. Inco Protocol is used to build an encrypted forum for anonymous employee feedback. Smart contracts are deployed on scalable blockchains like Polygon, Scroll, and Zircuit. Ledger Wallet is used for secure claiming of GTK tokens through hardware wallets. We are also integrating Nouns DAO which is a well-known community-owned brand that are unique, generative avatars bringing a vibrant and positive vibes. Nouns DAO is a useful public goods which embodies the spirit of open participation and it is implemented as the identity of the employees.

This project uses Chainlink to bridge real world data to the blockchain, through the chainlink external adapter which bridges the Iot device data to be called by the chainlink function job on the chainlink node hosted. This process is facilitated by a chainlink automation time-based trigger which is scheduled to run at a pre-specified interval, aimed at getting data from the IoT sensors and putting them on-chain after aggregation.

We also created an air quality sensor using an 8GB RAM Rasberry Pi, powerful enough to be able to run computations within the device on the edge. The device has 2 sensors, which captures humidity, temperature, CO2, NOx and other gases data.

For the private forum, it is important that the votes and messages remain anonymous, we have written a FHE-powered contract in the Inco Rivest Testnet that encrypts the votes and only allows decryption in some conditions. This protects the employees' identity from being exposed.

We leverage unsupervised AI learning with clustering method to classify companies' ESG risk level. In the future, data collected from the IoT sensors will contribute positively to our dataset and improve our models as they are re-trained. To avoid a single source of AI output, we leverage Chainlink oracle to aggregate different sources of AI models to ensure higher reliability and consistency.