2023 MIT IDE Annual Conference

• David Verrill – Executive Director, MIT IDE
• Sinan Aral – Professor, MIT Sloan and Director, MIT IDE

• Sinan Aral – Professor, MIT Sloan and Director, MIT IDE

In the Web3 group we identify, test, and validate designs and approaches to Web3 that ensure an inclusive and prosperous Internet future.

9:25am: “Decentralization of Blockchain Networks: Trends, Corollaries, and Causes”

• Harang Ju

Decentralization is a key feature of blockchain networks, as it allows for distributed trust, open protocols, and consensus among participants. However, decentralization can vary significantly across different blockchain networks, and there is little understanding of the historical trends, corollaries, and causes of decentralization. In this study, we propose to investigate these issues to provide a comprehensive and definitive basis for future studies of large-scale blockchain networks.We hope to provide valuable insights for researchers, policymakers, and industry practitioners on the evolution and governance of decentralized networks.

9:45am: “Information Rules Revisited”

• Madhav Kumar

The digital economy has undergone significant changes in recent years, particularly with the emergence of Web3 and its new economic primitives. As such, this research aims to examine the evolution of Information Rules, as outlined by Carl Shapiro and Hal Varian in 1999, and its adaptation to the changing economic landscape. By analyzing the impact of decentralized finance, non-fungible tokens, and blockchain-based marketplaces on these rules, this study aims to provide insights into the current state of the digital economy.

10:05am: “Scarcity and Valuation: Investigating the Impact of Limited Supply on NFT Pricing and Web3 Economics”

• Ehsan Valavi

The emergence of digital economies has led to new asset classes, such as non-fungible tokens (NFTs). While NFTs offer unique ownership and provenance characteristics, the role of scarcity in their pricing dynamics and broader Web3 economic implications has yet to be explored. This study proposes to examine the influence of scarcity on NFT pricing and the resulting effects on Web3 economic systems. We will utilize quantitative analysis of NFT transaction data to uncover the correlation between scarcity and price premiums in the NFT market while identifying potential moderating factors such as perceived utility, artistic value, and community support. Additionally, we will study how scarcity shapes market dynamics, including the emergence of NFT ‘whales’ and the potential for price manipulation.

10:25am: “Crafting Mechanisms and User Behavior in Blockchain Gaming: Insights from Axie Infinity”

• Harang Ju

Tokens are a fundamental medium of value in Web3, especially in gamified blockchain experiences. Emissions of tokens—from fungible “coins” to non-fungible items and characters—are essential to expand the supply or variety of tokens for a growing population of users in all applications of tokens. To support token value, supply can be balanced by allowing users to craft desirable digital items using existing tokens, both fungible and non-fungible.To date, how the design of crafting mechanisms influences users’ crafting behavior has not been explored. Here, we consider two crafting events in the blockchain game Axie Infinity, to examine their effects. We find that using any items to craft will increase the number of participants, but sequential crafting of rarer items increases individual usage. We suggest that system designers airdrop NFTs to users for maximum participation, and use design sequential crafting of rarer items to maximize retention.

Enjoy a short break

• Yamini Rangan – CEO, Hubspot

Keynote Address: The Power of Choice: Reflections on My Path to Leadership

Yamini Rangan has faced many twists and turns on her path to becoming a CEO. In this presentation, she will share what she’s learned about leadership throughout her career journey and why it’s important to make intentional choices. Her reflections carry advice and insight for anyone in leadership or aspiring to become a leader.

• Andrew McAfee – Co-director, MIT IDE

How can technological progress help humans flourish on a healthier planet? Andrew McAfee’s 2019 book, More From Less, shows that these two goals of human and technological growth are not incompatible; they’re inseparable. As societies become richer and more technologically sophisticated, they eventually decrease the burdens they place on the Earth. With additional support, the Tech for Good research group will pursue two new lines of research: How quantum computing and related technologies not only affect privacy and security, but simulate how nature works to lower production costs. Second, why some companies are so much more productive and innovative than others—and how to shrink that gap.

12:05pm: “The Business of Quantum Computing”

• Jonathan Ruane

Quantum Computing (QC) is in the early stages of transitioning from a laboratory curiosity into an engineering reality. Classical computing had a profound impact on our society, economy, and daily lives. Yet, important computing applications remain intractably out of reach, especially where complexity scales exponentially. QC offers a potential path forward but currently, it is a field that is fast evolving but with enormous challenges. Our objective is to shed light on initial use cases, industry sustainability, and near-term cryptography implications. Future work considers how we measure progress in the field from a commercial perspective– tying together technical developments with economic potential.

Enjoy a lunch break

• Neil Thompson – Research Group Lead, MIT IDE

“Quantum Economic Advantage: A New Practical Framework”

Discussions of quantum computing’s future are dominated by declarations of the number of qubits in new machines. But, like in classical computers, it is not just the hardware but also the algorithms that matter for what can be run and how fast. Achieving “Quantum economic advantage” will depend on both.  In this talk, I will present a practical framework for understanding which problems will be accelerated by quantum computing and which won’t.

1:20pm: “Can Specialized Computing Hardware Power Moore’s Law?”

• Tamay Besiroglu

Moore’s law has been a powerful and long-lived engine for progress in computing, generating breakthroughs in science and technology, and spurring economic growth. However, technology roadmaps have shifted to hardware specialization as the new engine of progress. And while specialization will be important, we are skeptical that specialization can produce similarly inclusive and sustained improvements. Using a simple model of hardware acceleration, we detail the range of technical and economic pre-conditions needed for specialization to power performance improvements. These include constraints about partial specialization, and the economic difficulty of financing R&D for specialized chips. Taken together, these arguments suggest that specialization will likely only be important for a subset of applications—and that even for these, specialization will be a bridge technology to post-CMOS technologies, not a long-standing era of its own.

1:40pm “Measuring Algorithm Contributions to Computing Performance”

• Jayson Lynch

While Moore’s Law has long been viewed as the driving force behind computing technology, software and algorithmic advancements have also played a significant role in computing performance. In this talk, we present findings from a survey of 113 important problems in computer science and over 1,000 algorithmic solutions to these problems. Our research estimates that more than 30% of these solutions have resulted from algorithmic improvements versus hardware improvements. Furthermore, we found that 65% of these problems now have asymptotically optimal algorithms, representing a significant achievement for the field but also restricting future improvements we can expect. By understanding the impact of algorithms on computing performance we can gain insights into future trends.

• David Rand – Research Group Lead, MIT IDE

Why do people share inaccurate and false news online and what can be done to reduce the spread? Cognitive science research and digital field experiments are tools to advance the fight against online misinformation. We shed light on the basic science of how people decide what to believe and share, and leverage these insights to design anti-misinformation interventions. Our experiments test the impact of these interventions, and collaborate with technology companies like Google and Meta to translate our science into real-world practices that drive accuracy in our digital lives.

2:20pm: “Quantifying the Impact of Anti-vaccine Information on Facebook”

• Jennifer Allen

Critics have attributed the low uptake of the COVID-19 vaccine in the U.S. to anti-vaccine misinformation spread on social media. We empirically evaluate the plausibility of those claims by quantifying the impact of anti-vaccine true and false content on Facebook. We combined randomized-survey estimates of the persuasive effects of 130 different pieces of vaccine content with exposure data from Facebook’s URL shares dataset. We find evidence that overall, exposure to COVID-19 vaccine misinformation lowers intentions to take a future vaccine. However, we also find that there is significant variation in treatment effects, and that this heterogeneity is predicted by the extent to which content suggests the vaccine is harmful to a person’s health. We further find that harm ratings of content, rather than veracity, best predict the treatment effect. We then use a novel method combining crowdsourcing, bayesian meta-analysis, and natural language processing to extrapolate the findings from our lab experiments to the larger universe of content on Facebook. We estimate that content identified as harmful by our model lowered vaccine intentions by approximately 1.5pp (CI: -2.28, -.328) per US Facebook user. However, this impact was largely driven by exposure to anti-vaccine content from mainstream news sites, rather than exposure to outright false content. Although fact-checked false and misleading content had a larger predicted negative effect on a per-view basis, it only constituted 2.1% of exposure to harmful vaccine content. Our results strongly suggest that platforms should not only consider the veracity of content, but also the relative harmful impact of that content when designing their policies.

2:40pm: “Using Twitter Ads to Shift Users’ Attention to Accuracy at Scale”

• Hause Lin

Prompting people to think about accuracy reduces the sharing of misinformation online. Here, we ask whether this approach can be implemented at scale using targeted digital advertisements encouraging users to think about accuracy before tweeting and sharing. Meta-analyzing multiple experiments in which we randomized whether users who had previously shared low-quality content on the platform were targeted with accuracy prompt ad campaigns, we find that the campaign did, indeed, reduce the number of low-quality URLs shared relative to the control condition. These findings suggest that shifting attention to accuracy is a promising approach for reducing the spread of misinformation online.

Enjoy a short break

• Dean Eckles – Research Group Lead, MIT IDE

In this research group we are developing, evaluating, and using new approaches to measure causal inference in complex data sets. We use massive digital experiments with corporate partners to better understand consumer behavior. Our goal is to enable rigorous discovery of insights from data, thereby informing business, public policy, and social sciences. We want to gain insights while preserving individual information. With additional support we will explore and design new methods of real-time data and analytics.

3:35pm: “Network Interventions to Reduce Hate Speech on Twitter”

• Eaman Jahani and Sam Fraiberger

As social media platforms become an increasingly inseparable part of our social structure, there are growing concerns about the prevalence of hate content on these platforms. Several social media companies have embarked on content moderation, through activities such as deleting content or banning users, as a way to combat online hate speech. Existing evidence on whether these punitive measures are effective at reducing hate speech are mixed. In contrast, community-driven interventions rely on the existing social capital in online networks and can potentially be more long-lasting. The goal of this randomized control trial-based study is to evaluate and compare the effectiveness of community-driven interventions aimed at reducing hate speech on Nigerian Twitter. We view online hate content as an ecosystem with supply and demand, and evaluate whether an intervention targeting hate speech producers is more effective than one targeting their consumers. In particular, the experiment enables us to compare the effectiveness of a producer-only vs a consumer-only targeting campaign while taking spillovers into consideration.

• Alex ‘Sandy’ Pentland – Research Group Lead, MIT IDE

In this research group we are studying how finance, manufacturing, data and human resources are being managed in a distributed world where the traditional roles of corporations and employees are facing profound transformations (e.g. DAOs). Current work is focused on how we can coordinate actors (employees and companies) to simultaneously optimize both efficiency and robustness in rapidly changing environments, the software and legal infrastructure that best supports distributed organizations, and the practical protocols that allow audit of supply chains for carbon tracking, fraud detection, and legal compliance. With additional support we will study how well these methods work in real environments in order to inform, guide, and shape critical decisions by industry, government, and society.

• Renée Richardson Gosline – Research Group Lead, MIT IDE

How can we understand human-AI interaction and shape systems that place humans first, encourage optimal decision making, and prioritize responsible innovation? The effective adoption of AI by employees, partners, and customers is becoming essential to competitive advantage. It is also critical to place the welfare of humans first, to maximize the benefits of human-AI collaboration, and to reduce harm. This research group examines human-AI interaction through a behavioral science lens, with a focus on human and AI symbiosis and how the balance between them varies by task. We study when humans can and should trust algorithms, how to prevent bias, and how to develop mutual learning. Using experiments and mixed-method approaches, the Human-First AI Group will proffer strategic, policy, and behavioral solutions.

4:35pm “Consumer Perception of Creative Contents Generated by Human Experts and AI (ChatGPT-4)”

• Jerry Zhang

With the rapid advancements in Large Language Models, there are four primary methods for generating creative content: (1) human-generated content, (2) AI-generated content, (3) human-edited content originally created by AI, and (4) AI-edited content initially created by humans. To investigate consumers’ perceptions of creative content produced under these four paradigms, we enlist ChatGPT-4 to create advertising materials for products and persuasive content for campaigns. Next, in collaboration with one of the world’s largest consulting firms, we engage its professional content creators to develop advertising materials for products and persuasive content for campaigns, both with and without the assistance of ChatGPT-4. We also ask ChatGPT-4 to edit content created by human experts without its aid. In a series of studies, we examine which content creation paradigm yields the highest quality. We also study the potential bias toward a specific content generation paradigm (e.g., how consumers’ perceptions of quality change when the creator’s identity is either revealed or concealed). The research sheds light on how to better promote Human-AI collaboration that delivers high quality output.

• John Horton – Research Group Lead, MIT IDE

In this research group we are identifying, designing, and evaluating problems, opportunities and solutions for the digitization of labor markets and work. Current projects include Measurement Tools for COVID-19 and Remote Work, Economics of the Sharing Economy, and the Future of Complex Crowd Work.

5:10pm “Algorithmic Writing Assistance on Jobseekers’ Resumes Increases Hires”

• Emma van Inwegen

There is a strong association between the quality of the writing in a resume for new labor market entrants and whether those entrants are ultimately hired. We show that this relationship is, at least partially, causal: a field experiment in an online labor market was conducted with nearly half a million jobseekers in which a treated group received algorithmic writing assistance. Treated jobseekers experienced an 8% increase in the probability of getting hired. Contrary to concerns that the assistance is taking away a valuable signal, we find no evidence that employers were less satisfied. We present a model in which better writing is not a signal of ability but helps employers ascertain ability, which rationalizes our findings.

5:30pm “B2B Platform Success – Lessons for Firms and Deviations from B2C”

• Leonardo Serra

Research on B2C platforms is well documented. In this research project, we are exploring research on the growth and vitality of B2B platforms. Drawing on data derived from more than 350 global B2B platforms, over 25 high-level executive interviews, and a qualitative survey with 200 participants, we seek to answer the question of how B2B companies can leverage the benefits of the platform economy. Our research reveals that B2B platforms vary significantly in functionality and can be differentiated into different types. As expected, key success dimensions differ widely from B2C platforms. Join us as we share our early findings and shed light on the identified key factors necessary for B2B platform success.

• Sinan Aral – Professor, MIT Sloan and Director, MIT IDE
• Andrew McAfee – Co-director, MIT IDE

Thank you for attending