# Semantic Information Marketing in The Metaverse: A Learning-Based Contract Theory Framework

**Authors:** Ismail Lotfi, Dusit Niyato, Sumei Sun, Dong In Kim, Xuemin Shen  
**arXiv:** [2302.11457](https://arxiv.org/abs/2302.11457)  
**Date:** 2023-02-22  
**Summary page:** https://www.summarizepaper.com/en/paper/2302.11457v2/

## Key points

- The paper addresses the problem of designing incentive mechanisms for a virtual service provider (VSP) to hire sensing IoT devices to sell their sensing data for creating and rendering the digital copy of the physical world in the Metaverse.
- Semantic extraction algorithms are proposed to reduce delivered data by sensing IoT devices due to limited bandwidth.
- Hiring sensing IoT devices to share their data with the VSP and deliver the constructed digital twin to Metaverse users is vulnerable to an adverse selection problem caused by information asymmetry between system entities.
- A novel iterative contract design is proposed using a new variant of multi-agent reinforcement learning (MARL) to solve this issue.
- Extensive simulations were conducted, and key performance metrics were measured for the Metaverse contract. Results show that the designed iterative contract incentivizes participants to interact truthfully, maximizing VSP profit with minimal individual rationality (IR) and incentive compatibility (IC) violation rates.
- The learning-based iterative contract framework has limited access to private information of participants, which is a first in addressing adverse selection in incentive mechanisms.
- A layer semantic communication system was also developed for the Metaverse followed by a reverse auction mechanism incentivizing data owners (sensing IoT devices) to make truthful bids selling semantic information to VSPs.
- The optimal pricing bundles can be derived if system designers have full knowledge about participant's private information while incentivizing them to reveal it truthfully.

## Summary

This paper addresses the problem of designing incentive mechanisms for a virtual service provider (VSP) to hire sensing IoT devices to sell their sensing data for creating and rendering the digital copy of the physical world in the Metaverse. Due to limited bandwidth, semantic extraction algorithms are proposed to reduce delivered data by sensing IoT devices. However, hiring sensing IoT devices to share their data with the VSP and deliver the constructed digital twin to Metaverse users is vulnerable to an adverse selection problem caused by information asymmetry between system entities. To solve this issue, a novel iterative contract design is proposed using a new variant of multi-agent reinforcement learning (MARL). Extensive simulations were conducted, and key performance metrics were measured for the Metaverse contract. Results show that the designed iterative contract incentivizes participants to interact truthfully, maximizing VSP profit with minimal individual rationality (IR) and incentive compatibility (IC) violation rates. Furthermore, this learning-based iterative contract framework has limited access to private information of participants, which is a first in addressing adverse selection in incentive mechanisms. A layer semantic communication system was also developed for the Metaverse followed by a reverse auction mechanism incentivizing data owners (sensing IoT devices) to make truthful bids selling semantic information to VSPs. The optimal pricing bundles can be derived if system designers have full knowledge about participant's private information while incentivizing them to reveal it truthfully.