Welcome

Welcome to the course blog for MS&E 135 Networks at Stanford University. This course explores how diverse social, economic, and technological systems are built up from connections, and how the study of networks can help us understand these systems.

Enrolled undergraduate students will be writing regular posts on varied subject matters and current events related to the course. Topics include: networked markets, social networks, information networks, the aggregate behavior of crowds, information diffusion, the implications of popular concepts such as “six degrees of separation” and the “friendship paradox.”

The blog is visible to the public, however only students and course staff are able to post and comment. Students should refer to the course blog guide for more information.

Future-Focused Algorithm Design: A Case Study in Ride-Sharing

Have you ever wondered how ride-sharing services like Uber match riders to drivers? This quarter, alongside Networks, I studied MS&E 112 (Graph & Combinatorial Optimization) with Dr. Amin Saberi, where we learned a lot about algorithm design. In our final class this week, Dr. Saberi spoke about his work in helping develop Uber’s matching algorithm. I am fascinated by these kinds of matching problems, and as a possibly too-avid Uber customer, I found this topic particularly interesting.

Though I am still reading through the paper Dr. Saberi co-authored on this topic and seeking an understanding of the more technical aspects, I wanted to use this blog post to discuss the higher-level objective and methodology of the algorithm, as well as its implications in the real world.

Most particularly, I want to focus on contextualizing this algorithm by examining the idea of a future-focused algorithm more broadly. How can algorithms like this help in areas beyond ride-hailing?

Uber Algorithm Overview

Essentially, we want to find the optimal matching of riders and drivers. Given a static, bipartite graph where we have each rider and driver represented as a node and connected by edges, we could do this by finding the maximum matching. However, the key caveat in ride-sharing is that riders do not all arrive at the same time. They arrive one by one, and they need to be matched relatively immediately with a driver. A real-time matching algorithm like this is known as an “online” algorithm, and it introduces a level of complexity that we don’t see in “offline” algorithms where the nodes and edges are already known.

One key theme in the Uber algorithm is its consideration of the future. For example, to find the optimal matching of riders and drivers, the most obvious approach would be to match each rider with the driver closest to them the moment they request a ride. However, this would not be optimal, since the next rider may end up having to be matched with a driver much farther away. Had the algorithm waited a little while and considered both riders together, a more optimal matching could have been reached. Thus, the Uber algorithm uses the concept of “batching” to wait a little and let several riders and drivers accumulate in the network, then match them all at once.

This future-focused approach also appears in another aspect of the algorithm. What would happen if the algorithm matched a batch of riders with drivers, and then all of the drivers in a certain area were occupied? In the next batch, riders in that area would have to be matched with drivers who are much farther away. To address this, the Uber algorithm does not just consider the riders who have already requested rides, but it also considers riders who are likely to request a ride in the near future. To do this, it assigns each potential rider (users who have opened the app) a probability that they will end up requesting a ride. The matching will then take these riders into account.

Contextualizing this Algorithm: Applications of Future-Focused Algorithms Beyond Ride-Hailing

This algorithm improved the optimality of the matchings significantly. So to broaden our perspective and contextualize this algorithm, we can consider the following question: How can we apply this idea of a future-focused algorithm in areas beyond ride-sharing?

The applications of incorporating information or probabilities about the future state of the graph into algorithmic design are endless. For a very basic hypothetical example, let’s say Starbucks could note when people open the app and assign a probability that they will order a particular item. Then, aggregating this data could possibly help managers and employees decide what ingredients and blends to prep.

Key Takeaway

When network-based algorithms can incorporate information about the future into the network, they can optimize much more effectively. This is a fascinating topic and one that I hope to explore more in the future!

Sources

Y. Feng, R. Niazadeh, A. Saberi, Two-Stage Stochastic Matching and Pricing with Applications to Ride Hailing, Operations Research, 2023.

Real Estate and Auction Theory

The real estate market is constantly adapting, and is now accelerating with advancements in technology. Traditional ways of buying and selling homes are being challenged by new ideas like iBuying and auction theory.

According to a recent Forbes article, auction theory, a branch of economics that studies bidding strategies and outcomes in auctions, could revolutionize how homes are bought and sold by introducing more transparency and competition into the process.

Auction Theory in Action

Auction theory, similar to the Nash equilibrium in game theory, suggests there are best moves for buyers and sellers. For real estate, this could mean sellers getting the best price while buyers find homes within their budget. Designed right, auctions can balance these needs, leading to better deals for everyone.

Auction theory operates on the premise that there are optimal strategies for both buyers and sellers in a market. In real estate, this could mean creating an environment where sellers can maximize their property’s value while buyers have a fair chance at purchasing homes within their budget. Auctions, especially those designed with careful consideration of auction theory, can help achieve a balance between these two objectives, leading to greater satisfaction for all parties.

Strategies for Buyers and Sellers

Grasping the basics of auction theory can give buyers and sellers a strategic edge in the real estate market. For sellers, it means structuring the sale process to attract more buyers and encourage higher bids. For buyers, it means understanding the dynamics of bidding strategies to make competitive offers without overpaying.

Looking at auction theory and real estate through the lens of game theory shows us new ways to innovate and improve. These models not only help individuals take advantage of the market, but also encourages us to deeply reflect on the decisions of others.

Source: Forbes Real Estate Council

Diffusion of Ideological Innovation: A Network-Analytical Understanding of the Protestant Reformation

History has traditionally been interpreted through qualitative methods, relying more on narrative and less on hard data. This field has been somewhat wary of quantitative approaches and formal models, but it’s precisely this hesitation that piques my interest—there’s a sense of uncharted territory in applying these analytical methods to historical study. I feel that these imprecise, qualitative methods fundamentally generate hypotheses. We must actually test these hypotheses using rigorous methods, like network analysis, subject to available data for and information regarding the time period and subject. It is particularly exciting to see conclusions derived from these methods fit into the existing historiographical landscapes. Hopefully, they provide a novel perspective on well-trodden subjects and can transform our understanding of historical events.

Our exploration of the diffusion of innovations prompted me to consider not only technological innovations, but also those intellectual and ideological. Specifically, I hoped to explore the how the novel ideas of the Protestant Reformation diffused throughout Europe. The source I read, Becker et al., specifically examined Martin Luther’s central role in the Reformation through advanced network analysis and simulation techniques. They draw on foundational concepts from network analysis in order to formally model the history, like nodes (cities) and ties (Luther’s correspondence, visits, and student relationships), and explore the dynamic process of ideological spread similar to both the diffusion of technological innovation and spread of disease as we explored through the epidemiology portion of the class. However, their (Becker et al.’s) exploration leverages the notion of “multiplex networks,” which refers to the “superposition of several networks on the same set of nodes” (859). In this particular context, they analyzed two distinct networks: the network of Luther’s influence (e.g., Luther’s correspondence, visits, and student relationships) and the network of trade throughout Europe at the time.

Their study also used network simulations in which they tested various diffusion scenarios to determine how Luther’s influence, spatial diffusion via trade routes, and their interplay contributed to the spread of the Reformation. This allowed them to “turn on” or “turn off” certain network ties or effects to observe potential outcomes. These simulations also relate to the “epidemiological models” we studied in class. For example, they employed probabilistic rule-components for nodes becoming infected. They also use some concepts from the non-probabilistic model from class on diffusion of technological innovations. For example, in simulations they employed rules like nodes adopting the technology once a certain threshold proportion of neighbor/connected nodes had already adopted.

Ultimately­—and this is where the “multiplex” concept comes in—they find that only when they tested the multiplex superimposed network did the simulation actually show successful diffusion. With testing of the two separate networks, diffusion was not successful as seen in historical reality. Thus, they conclude that both factors—Luther’s personal influence and network, as well as the macro circumstances of trade networks—were essential to the success of the Protestant Reformation spreading throughout Europe.

Source: https://journals.sagepub.com/doi/pdf/10.1177/0003122420948059

Private Equity Games: Investments & Carried-Interest Waterfalls as Principal–Agent Contracts

For me, one of the important takeaways from the class was how behavior in response to private signals can be used by the public to infer about that private information; it is a way to try to mitigate asymmetric information. In the private financial markets, two fundamental relationships characterize a game-theoretic analysis: (1) investment into a private company/asset (e.g., startup) by an investor (e.g., PE fund), and (2) the limited partnership agreement (LPA) between the General Partner (private equity fund) and Limited Partner (client, e.g., pension, endowment fund, life-insurance).

Both relationships exemplify different principal–agent models. In the first (asset-level investment), the PE fund is the principal and the asset’s (company’s) management team is the agent. In the second relationship (fund-level), the LP contracts out the “work” of investing to the GP. Thus, we can conceive of the market of private equities as modeled by a layer of fund-level LPA principal–agent contracts stacked on top of a layer of asset-level principal–agent contracts.

The concept of carried interest in a limited partnership agreement within the context of private equity can be analyzed through the lens of the principal-agent model using game theory, offering a structured way to understand the dynamics and incentives at play. In this model, the principals (limited partners or LPs) invest capital in a fund managed by agents (the general partner or GP), with the expectation that the GP will act in the best interest of the fund, thereby maximizing returns for all parties.

The two sources I read discuss the effectiveness of standard/conventional investment contracts (in both the asset– and fund-level relationships) at efficiently aligning the incentives of the principal and the agent, relating to the normative question of how we should structure “the game” between principals and agents in private equity (i.e., do the currently conventional contract structures ought to change?).

For the sake of space, I will focus on the fund-level, LPA model, for which the conventional contract is the carried-interest waterfall structure. Here, the principal–agent problem arises from the inherent conflict of interest between the principal (LPs) and the agent (GP). The GP has better information on its own “skill” at investing than the LP; this is asymmetric/hidden information, or adverse selection. Furthermore, the LP has no inherent way of ensuring the GP applies sufficient “effort”; this is known as hidden action/moral hazard. The LPs aim for the highest possible return on their investment, whereas the GP seeks to maximize its own wealth, which may not always align with the LPs’ goals. This misalignment can lead to issues such as the GP taking excessive risks (type of moral hazard).

Carried interest simply refers to the compensation structure as paid out to the GP and LPs based on realized investment returns. Specifically carried interest represents a share of the profits earned by the private equity fund that is paid to the GP, typically around 20%, after achieving a certain return threshold or “hurdle rate.” This mechanism aligns the interests of the GP with those of the LPs by directly tying the GP’s compensation to the performance of the fund. The better the fund performs, the more the GP earns in carried interest.

The sources discuss how carried interest is mostly effective at aligning incentives. Profit-sharing, for example, helps address the information asymmetry by allowing competent GPs to signal confidence in their abilities to the market of LP capital. LPs can interpret/infer the GP’s private information (about competence) from that signal, mitigating adverse selection. Profit-sharing also helps mitigate the moral hazard of the LPs’ inability to monitor GP effort/choice of investment strategy because profit-sharing is performance-based compensation. Furthermore, carried-interest structures often contain return-of-capital/preferred hurdle tiers that prioritize the LP, returning capital to the client before the GP sees any dollars; this “first-loss” component protects the LPs downside-risk by reallocating the risk of underperformance to the GP.

However, the carried-interest structure is limited in aligning incentives. Mehta notes how you can view it as a call option on the value of the investment vehicle/fund. This essentially comes from the risk structure, as the downside risk is limited (strike price analogous to initial GP capital contributed), while there is massive upside. In addition, the fixed, fee-based compensation helps mitigate the downside risk. Viewed as a call option, we can see that increased risk-taking (volatility) will increase the value of the option (which represents GP’s carried interest). One “solution” I see is that these investment games don’t occur in separate vacuums. In reality, these are repeated games in the larger context of an ongoing fund manager–LP client relationship. As with repeated games, reputation now becomes an important factor, mitigating both information asymmetry and the moral-hazard of excessive risk-taking.

In addition to repeated games, further additions to the model could be increased granularity, where nodes are individuals and not entities like investment firms. For example, there’s employee-level compensation, where individuals get points of carried interest in the fund on top of a salary/bonus. If someone is about to retire, reputation in the repeated game theoretically does not matter, which can misalign incentives and create a moral hazard of excessive risk-taking in regards to the employee’s investment decisions on behalf of the firm.

These are important questions: these principal-agent contracts allow for specialization of skills. LP clients focused on diversifying according to their investment goals for endowments, pension funds, GPs can focus developing particular skills in private equity investing, or get even more granular like focusing on a specific industry or vertical in biotech specifically. We can think of these principal-agent contracts as edges linking the nodes of different kinds of investors to companies/investments. Trillions of dollars of private capital flows through the edges in this network. It is thus important that these principal-agent contract edges function efficiently. Whether in a pension fund, or has a high-net-worth individual invested in a VC fund, your money will likely be at stake in this principal-agent network.

Sources:

Vijay Mehta: “Principal-Agent Issues in Private Equity and Venture Capital” (https://core.ac.uk/download/pdf/76379595.pdf)

Ludovic Phalippou: “Modifying the Carried Interest to Do What It Is Said to Do” (https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3333053)

Kate Middleton and Information Cascades

https://www.cnn.com/2024/03/15/world/kate-middleton-news-controversy-cec/index.html

After not being seen in public for some time, an edited photo of Princess Kate Middleton with her children paired with the excuses of the Palace has led to a flurry of questions and conspiracy theories.

This is information cascades at work. Note that I do not intend to take any stance on what is true or not about the situation. However, many people would simply not have noticed her absence at all, would not have thought twice about the Palace’s excuses, or would not have noticed the botched editing. But the more people were saying that the situation was suspicious, the more people believed that it was suspicious. How could there be nothing there if so many people are talking about it?

As with all cascades, people are acting based on the actions of others, started with very little information.

Information cascades and their relevance in today’s age of digital manipulation

Informational cascades are at the forefront when it comes to influencing perception, particularly in the face of AI and social media. They form most often when individuals are fed information, whether erroneous or valid, and the public interprets and reacts based solely on the dissemination of the information. This was evidenced this week as the media speculated on the whereabouts of Princess Catherine and the public continued to seek answers. The royal family has been quiet, leaving perhaps even more room for conjecture. We were told that Princess Catherine was recovering from a surgery and has been out of the public domain for weeks. Earlier this week, the Princess posted a photograph of herself and her family, allegedly taken by her husband, Prince William. It seems the photograph was released in order to dispel rumors that have been swirling regarding Princess Kate’s absence from the public view. However, the photograph has stirred up even more suspicions. It was analyzed on social media and by the press which raised questions as to its authenticity. Discrepancies were noted and Princess Catherine ultimately confessed to manipulating the photograph without an explanation as to her motives… 

The use of AI and digital manipulation has resulted in lots of skepticism in recent times. Can we rely on digital media to tell a story or has it become misleading, rendering it useless? 

This news story highlights the fundamental concept that information cascades can just be straight up wrong and erroneous. Although the statement explicitly said that Middleton was recovering from surgery, the public paid no attention to this at all, focusing solely on the image, and extrapolating that to question the validity and honesty of the royal family. It appears that there is often no explanation for some aspects of human psychology. Humans enjoy concepts that have a lot of buzz around them, and there seems to be this innate human desire for instigation. I can hardly explain it in a more succinct manner than that.  

Furthermore, in order for information cascades to thrive, they require dissemination through mass distribution on platforms such as social media. Information is readily available through these avenues for public consumption and analysis. The photograph posted by Princess Catherine was widely distributed and unleashed a frenzy of commentary as to its authenticity and the circumstances surrounding the Princess’ absence. The cascade of information also set in motion a battle of opinions and explanations which essentially took on a life of their own. In this right, information cascades demonstrate power in persuasion and influence. Individuals suggesting that the photo was edited not only raised more questions as to if, in fact, it was but also as to why, leading to all sorts of theories on Princess Catherine’s health, marriage, and venue. The public “jumped on the bandwagon” inferring that the royal family must be hiding something.

Clearly in the age of AI, and particularly as its tools get sharper, prudent individuals may opt to limit their reliance on information cascades. Those who are prudent may choose to use a variety of sources and further investigation before making assumptions. Social media may be discarded as an authoritative source and instead, used just as a platform for discussion with individuals looking further for direct evidence. Perhaps the photograph that was posted by Princess Catherine was edited solely because she was recovering and didn’t look her best? Inquiring minds through information cascades and their inflammatory effects have led inquiring minds to all sorts of conjecture here. So until we have direct evidence, possibly a statement from the royal family or the Princess herself, the public will continue to consume what social media outlets feed. 

There are so many ways in which we can be influenced by others in society nowadays. We are influenced by others to buy certain products, hold certain political positions, and imitate their choices in order to maintain a status quo. 

The Kate Middleton example draws great comparisons to some of the phenomena surrounding information cascades that we study in class. We have already discussed the imitation in decision making that occurs through social media platforms. We also discussed the social pressures and conformity with regards to the scandal in that with very limited information, skepticism gains momentum and individuals are essentially forced to believe the falsity in photography that is being forced upon them. It is a known fact that cascades are “based on little information”. From the text on information cascades,  “individuals in a cascade are imitating the behavior of others, but it is not mindless imitation. Rather, it is the result of drawing rational inferences from limited information.” In Kate Middleton’s case, I truly believe there was so much leverage and backing by the public that these images were real that the only “rational inferences” that followers could make would align explicitly with what their peers before them had thought. One interesting inference that I made after learning about Kate Middleton’s story was that while there was an abundance of informational effects meaning “ where individuals infer information from others’ actions”, there was no true existence of direct-benefit effects, or “where individuals align skepticism with prevailing sentiment to avoid social repercussions”. Through analyzing the cascade of false information that trickled through social media, it seemed to me that people were nonchalantly and inadvertently becoming followers. There was no true malicious intent in believing that these images were fake. Nobody was attempting to fulfill a personal agenda, yet the need to conform to what everyone else thought still felt necessary. I would really like to dive more into direct-benefit effects in the future to learn in what avenues they exist and how information cascades lead to tangible benefits for subjects of said cascades. 

Overall, the digital manipulation evident in this real world example highlights the need to be cautious when looking at things on social media and subsequently, assuming our peers’ opinions as our own. It is never smart to draw conclusions solely from the behavior of a crowd, whilst ignoring your own signals and intuitions. Information cascading effects display the need for critical thinking, especially in an age when misinformation, including manipulated photography, is so prevalent.  We are all new to this season of AI and its dissemination across social media. Yet, this incident with Kate Middleton reminds us to actually challenge our own intuitions and to take a second look at judging the validity of something that we may have never batted an eye at years ago. 

https://www.fastcompany.com/91056450/the-kate-middleton-scandal-shows-the-worst-of-a-post-generative-ai-social-media-world

https://www.newyorker.com/culture/annals-of-appearances/the-royal-photo-that-was-too-good-to-be-true

The Birth of Urban Air Taxis: A Less Congested World

How Air Taxis Will Work | HowStuffWorks
Computer visualization of urban air taxis (HowStuffWorks)

If you have ever driven through San Francisco, LA, or any major city, you have likely experienced some of the negative effects of traffic congestion firsthand. Urban traffic comes with a wide range of concerns: long commute times, high levels of noise pollution, and increased air pollution, to name only a few. To solve this problem in the long term, some recent startups are beginning to ponder the idea of urban “air taxi” networks.

For example, venture-backed Joby Aviation has been developing an eVTOL (electric vertical take-off and landing) aircraft that it intends to use as part of an urban air taxi fleet. The company has already successfully tested its prototype in Manhattan, and it plans to launch its services in Dubai by 2026. It has also passed 3 of 5 FAA certification stages, meaning that it may begin commercializing in America very soon.

Joby's electric air taxi is one step closer to taking flight : NPR
Joby’s prototype eVTOL (NPR)

Clearly, Joby’s technology is not ready for mass adoption, since it will rely heavily on infrastructure (e.g. small urban airports) that is not yet widespread. However, the company is working towards this by partnering with companies such as Skyports Infrastructure, a firm that aims to develop and design small eVTOL “vertiports.” Once the necessary infrastructure is in place, it is entirely possible that air taxis will become a part of everyday urban life.

Skyports Infrastructure's design for vertiports approved in Dubai | Skyports  Infrastructure
Dubai vertiport design (Skyports)

As a pilot myself, I find this idea very fascinating, and I want to consider the effects that a shift like this could have on travel. Framing my thinking using networks has provided me with some key insights. In this blog post, I want to focus on one area in particular: the effects on traffic congestion.

From 2D to 3D: More Options, Less Congestion

Moving traffic from a 2D plane – roads – to the 3D world of the sky would spell a significant reduction in traffic congestion. Firstly, the concept of traffic congestion that we studied in class relies on the basic assumption of fixed routes. This is a reasonable assumption, because in the short term, road networks are relatively immutable. There is a limited amount of infrastructure in existence, both in terms of the layout of the roads as well as their capacities. If we envision a road network as a graph, we can think of cities/areas/places as nodes, roads as edges between these nodes, and road dimensions (length and width) as proportionate to the capacities of each edge.

This framework drastically changes when we consider air travel. When we introduce a “z-axis,” the idea of a fixed network of routes loses its validity. As traffic gets heavier, every individual flight route can be easily adjusted by the pilot or by air traffic control operators in order to accommodate this increase. These adjustments could materialize as variations in altitude, heading, and so on. With the possibility to divert traffic along alternate routes, it is safe to say that there will be much less congestion.

Predicting Congestion in an Urban Air Taxi System: A Challenging Task

One interesting question that arises is: How can we model and quantify this potential decrease in congestion? Starting at the most basic level, how do we design a network that models an urban air taxi system? We can consider the airports or “skyports” as nodes. But what about the edge capacities? These cannot be as easily defined. Trying to define them will require an interdisciplinary approach, combining physical limitations (e.g. maximum flight altitude), airspace regulations (e.g. the airspace surrounding major airports and military bases is usually restricted to some degree), and airport infrastructure (e.g. only a certain amount of traffic can take off or land per unit of time). Furthermore, flight routes can be subject to change based on factors, such as wind angles, that do not typically affect road routes.

And beyond simply defining the network, what about analyzing it? If we want to understand traffic congestion, we will need a different approach for air routes than for road routes. This is because applying the concepts of Nash equilibria and payoff matrices becomes much more complex when the routes are not fixed (i.e. we do not have fixed “strategies”). Performing this kind of analysis would likely be rigorous and is beyond the scope of my current understanding, but I definitely plan to look into it more.

Conclusion

We have briefly delved into the effects that urban air taxis could have on traffic congestion. At a high level, we can see that an urban air taxi network would be less congested than a road network, but quantifying this result is more difficult. All in all, this is a fascinating topic with a lot of nuance! What’s more, it is highly relevant to our own lives in the future. Maybe in 10 years, calling an air taxi will be as easy as calling an Uber. The network effects of such a system will no doubt be profoundly widespread.

Sources

https://www.jobyaviation.com/

https://skyports.net/skyports-rta-and-joby-to-launch-air-taxi-service-in-dubai/

Mobility Data Enables Impressively Accurate Predictive Network Models

Guan et al (2021) is a study that was conducted during the height of the Covid-19 pandemic, and focused on designing a predictive network model using cell phone mobility data from Israel to estimate where and when disease outbreaks were likely to occur. With the vast store of mobility data at hand, the researchers were able to simulate future behavior of large groups of people with respect to the interactions that they are likely to experience through the social network on a given day. Combining this with the particular viral characteristics of Covid-19 produced accurate estimates for outbreaks that would occur in the near future.

The results of this study are interesting to consider as an intersection between factors of how social behaviors develop and how diseases spread. They can be quite similar, such as in how they require people to come together in order for development to occur, but can also be quite different, as in how social behaviors may hinge on obscure factors with blurry definitions while diseases have more predictable, quantifiable parameters. 

The implications for the potential applications for mobility data here are also massive. While many people tend to prefer less tracking of their mobile devices for privacy purposes, anonymized use of this kind of data can produce incredibly predictively powerful network models.

Source: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0253865

The Cultural Power of Power Law Curve

The power law distribution is a probability distribution in which there are many smaller values observed and only a few large values observed. This type of distribution is often used to describe the “rich get richer” phenomenon, but it can also be applied to the concept of “popularity” in a number of different contexts. In this way, the power law distribution becomes highly relevant to social networks and cultural phenomena.

In the world of entertainment, there are numerous examples of the influence of the power law distribution. Analysis of the distribution graphs for values such as box office earnings, Netflix series views, and Spotify streams reveal a stark commonality: the same power law curve. This provides some insight into how entertainment like movies, shows, and music become popular. The influence of behavioral conformity in networks, taking the form of social pressure, likely has a significant effect on what people choose to engage with, perhaps even more than the pure value of the options.



Another way to observe the influence of the power law distribution is in social media. Social media is highly similar to the types of entertainment discussed before in that people must choose what to spend their time engaging with. The display of follower and view counts on social media accounts and posts is a direct social signal of popularity that influences the behavior of some people who likely would not take notice otherwise. A good example of this is displayed with the graph of the percentage of subscribers, or patrons, held by the top content creators on the subscription-based content platform, Patreon. Once again, we see the power law curve, where just a few of the top creators hold substantial percentages of the total patrons. This is very reminiscent of the discussion of a relatively tiny amount of people in the United States owning the majority of the money.



In conclusion, the power law distribution is a force that cannot be ignored in the study of preferential decision making and the cultural phenomena that result.

Source: https://dougshapiro.medium.com/power-laws-in-culture-27ab6461c693

Decoding the Dynamics of Information Networks and Cascade Effects in Society


The digital era has revolutionized the way information is disseminated, leading to the rapid spread of ideas and behaviors across global networks. This phenomenon, known as the information cascade effect, occurs when individuals, influenced by the actions of others, adopt behaviors or beliefs en masse, often sidelining their personal information or judgment. This effect, while inherently based on rational decisions, can sometimes lead to widespread acceptance of misleading or false information, as seen in the proliferation of fake news.

Information cascades are not limited to the digital world; they are part of the fabric of human interaction. Whether choosing a restaurant, buying a product, or endorsing a political view, people often rely on the choices made by others before them. This behavior is rationalized by the belief that others’ actions reflect valuable information, especially when personal knowledge is limited or ambiguous. For instance, the preference for a busy restaurant over an empty one on the assumption that the crowd knows best exemplifies an information cascade in action.

However, the cascade effect is not just a mindless following of the majority. It stems from rational inferences made under limited information. The real estate market has long been viewed as a cornerstone of sound investment, often associated with stability and steady growth. This perception, deeply ingrained in the societal psyche, has led many to view property investment as a safe haven, particularly during times of economic uncertainty. However, the dynamics of information networks and the cascade effect have played pivotal roles in transforming these individual investment decisions into a collective movement, culminating in the real estate bubble and its eventual burst.

Initially, the allure of real estate as a lucrative investment spurred individuals and investors to enter the market, driven by the belief that property values would continue to rise indefinitely. This belief was reinforced by the prevalent narrative within information networks, where success stories of real estate investments amplified the perception of its profitability. As more people observed the apparent financial gains in the real estate sector, the information cascade effect took hold, leading to a surge in demand for properties.

This increased demand, fueled by optimistic speculation and the desire to capitalize on rising property values, drove real estate prices to unsustainable levels. The information network surrounding real estate investment became self-reinforcing, with each new buyer’s success story encouraging more individuals to invest, further inflating the bubble. Financial institutions, responding to the booming market, relaxed lending standards and introduced innovative but risky mortgage products to accommodate the growing demand, thereby exacerbating the situation.

However, the real estate market’s reliance on continuously rising property values and the ability of borrowers to refinance or sell at higher prices was inherently fragile. When the economic conditions shifted, exposing the speculative nature of the market, the bubble burst. The collapse was precipitated by a confluence of factors, including rising interest rates, over-leveraging, and a saturation of the market with properties that could not be sold at anticipated high prices.

The information network that once fueled the real estate boom turned into a vector for panic and rapid decline. As the reality of the market’s instability became apparent, the same cascade effect that drove the market up accelerated its fall. Investors and homeowners rushed to sell properties, financial institutions faced unprecedented defaults, and the once buoyant real estate market became the epicenter of a financial crisis.

The collective behavior of market participants, driven by shared beliefs and reinforced through social and financial networks, can create conditions ripe for economic upheavals. Understanding these informational dynamics is crucial for developing strategies to mitigate the risks associated with such speculative bubbles in the future. In conclusion, the dynamics of information networks and cascade effects play a pivotal role in shaping societal behaviors and decisions. By comprehensively analyzing these patterns, individuals and policymakers can better navigate the complexities of the digital age, ensuring more informed and balanced outcomes in various spheres of life.

Sources:

Source 1: https://www.americanprogress.org/article/2008-housing-crisis/

Source 2:

https://www.economist.com/finance-and-economics/2018/04/05/fake-news-flourishes-when-partisan-audiences-crave-it

Apple’s Crossroads: Power, Innovation, and the Richer-Get-Richer Dynamic in the Tech Ecosystem

In the intricate dance of power and economics within digital and social networks, the case of Apple Inc. offers a compelling narrative that echoes broader principles outlined in sociological and network theory research. This analysis draws upon both recent developments around Apple’s challenges and strategies, and theoretical insights from network analysis, particularly focusing on how network positions influence power dynamics and economic outcomes. The synthesis of these perspectives sheds light on the current state of the tech industry and offers predictions about its future direction.

The theoretical framework suggests that power within networks is not merely a function of individual attributes but significantly influenced by one’s position within the network’s structure. This is particularly relevant in understanding Apple’s situation, where its “walled garden” approach has been both a source of strength and a point of contention. Apple’s dominant position in the tech ecosystem allows it to exert considerable influence over developers, competitors, and regulators. However, recent challenges, such as the EU’s antitrust actions and competition from companies like Huawei, highlight the dynamic nature of power and the ongoing negotiation between maintaining control and fostering innovation.

The concept of power as a relation between entities, rather than an inherent attribute of an individual, is vividly illustrated in Apple’s interactions within its ecosystem. The company’s ability to dictate terms, control market access, and set standards showcases how power dynamics play out in economic transactions and social interactions mediated by technological networks. Moreover, Apple’s strategic maneuvers, such as its focus on developing generative AI technologies and potentially revolutionary products like the Vision Pro, can be viewed through the lens of seeking to maintain or enhance its position within the network.

In the broader context of network theory, Apple’s challenges and strategies also reflect the tension between navigational and transactional links within the information network of the web. As the web evolves, the balance between facilitating navigation and enabling transactions has become a critical aspect of how companies like Apple structure their interactions with users and developers. This duality is akin to the theoretical distinction between stable, public content and the more dynamic, transactional exchanges that characterize modern digital platforms.

Understanding Apple’s position and strategies in light of network theory and the sociology of power offers valuable insights into the tech industry’s future. It highlights the importance of network structure in determining economic and social outcomes, the dynamic nature of power, and the ongoing negotiation between different stakeholders within the network. As the digital landscape continues to evolve, the interplay between network position, power dynamics, and strategic innovation will remain central to shaping the trajectories of leading tech companies and their impact on the broader economy and society.

This blend of theoretical insight and practical analysis underscores the complexity of navigating the tech industry’s power dynamics. It invites further exploration into how companies can leverage their network positions to innovate and thrive, while also navigating the ethical and regulatory challenges that arise in an increasingly interconnected world. Therefore, it should come as no surprise that Apple’s decision to gradually adopt generative AI is prudent.

Source:

https://www.economist.com/business/2024/03/03/apple-is-right-not-to-rush-headlong-into-generative-ai

A Pending TikTok Ban: The Hidden Network Effects

TikTok, known for its short video snippets and endless scrolling, has certainly captured the zeitgeist of our digital era. Yet, beneath this popularity, a storm brews: American lawmakers, eyeing the platform’s Chinese roots with suspicion, have sounded the alarm over potential breaches in U.S. citizens’ data who are users on the platform. This imminent ban has led to fervent debates, especially amongst its millions of U.S. users who love and cherish the platform, however this paper will delve into the impacts such an action could unleash upon TikTok’s vast social network and the other titans of social media.

At the heart of TikTok lies a steady recipe of growth: users have the ability to seamlessly morph into creators, who quickly generate more users for the platform. Such a ban, however, would trigger a cascade effect, with a mass exodus of its 150 million U.S. users and creators suddenly having to search for a new digital “home” to make and consume content (Social Media Today). This could amplify the user base of rival platforms that have similar content layouts (think, YouTube Shorts and Instagram Reels), which would certainly alter the balance of power in the greater social media realm. The redistribution of TikTok’s massive U.S. user base could also intensify competition among existing platforms, pushing them to innovate more aggressively to capture and retain TikTok’s fleeing users. This competition could lead to rapid developments in technology, user experience, and monetization strategies, which would also reshape the social media landscape in unexpected ways.

But the ramifications of a TikTok ban also pierce deeper. At their core, they demonstrate how a single government act in the digital world not only impacts the targeted group, being TikTok and its users in this case, but also the myriad of other platforms and services entangled in our digital web. In essence, the looming shadow of a TikTok ban is more than Gen Z vs. Congress regulatory drama; it’s a stark portrayal of the unpredictable, often tumultuous waves of change that can occur if a big player in our social media consumption is removed. It also urges us to continue looking for the fine line between government regulation and innovation.

Sources:

Social Media Today: https://www.socialmediatoday.com/news/new-report-looks-tiktok-usage-behaviors-the-us/708275/#:~:text=TikTok%2C%20as%20noted%2C%20has%20over,among%20younger%20cohorts%20in%20particular.

Cumulative Advantage in Sports

The rich-get-richer phenomenon generally describes the phenomenon in which those who start out with an advantage in a field are placed in a position that makes them more likely to gain further advantages and opportunities, resulting in an increasingly widening gap between those who started with and without the advantage. This concept is also called cumulative status bias, or the Matthew Effect.

This paper explores this concept in the context of NBA All-Star elections, asking the question of how cumulative status bias manifests itself in these cases. The NBA All Star game typically takes place mid-season, designed to highlight talented players in the league. In other words, it is a significant event where players can gain status in the league in addition to benefits such as financial bonuses.

In following a sample of 1829 players, the paper found that only 172 players were ever named to the All-Star roster, and that among these 172 players, 111 players received more than 1 nomination. More specific calculations revelaed that while players who had never been nominated had a 3% chance of being elected to the roster, players who had been nominated previously had a 63% chance of being reelected—a classic example of where an initial “advantage”, this case being elected on the All Star roster, leads to further advantages down the line.

This poses an interesting question regarding what is going on in the judge’s minds when they make the decisions to result in such statistics. How much of the decision is subconscious? Can one count it as a subconscious decision of a judge actively weighs in players’ previous playing/election histories when determining who to elect in the new cycle? And if this is the case, is the rich-get-richer phenomenon unavoidable if it is impossible to evaluate the skill of a player without considering their previous accolades?

https://journals.sagepub.com/doi/full/10.1177/00031224231159139

Influencer Marketing : Echoes of the Get Rich-Get Richer Model on the Web

In the digital marketing landscape, the rise of influencer marketing has been nothing short of rapid. This marketing strategy allows companies to collaborate with social media influencers to promote products, brands, or services. It has become a staple for businesses aiming to enhance their online presence. However, beneath this innovative marketing approach lies a parallel to a much-discussed digital age phenomenon: the rich-get-richer model, also known as the Matthew Effect or preferential attachment.

Initially observed in the context of web page links, the get-rich-get-richer model describes how popular nodes (websites) tend to accumulate more connections over time, thereby becoming even more prominent. This pattern is observed in various domains, from wealth distribution in economies to the popularity of specific network nodes.

When it comes to influencer marketing, a strikingly similar pattern emerges. Large companies tend to gravitate towards influencers with already substantial followings to maximize the impact of their marketing campaign. This preference is based on a simple premise. Influencers with more followers can potentially reach a larger audience, thus offering a higher return on investment for the marketing effort.

Large companies’ selection process for influencers closely mirrors the rich-get-richer model. Influencers who have already achieved fame or notoriety find themselves in a virtuous cycle, where their popularity makes them more attractive to businesses. As these influencers are chosen for more campaigns, their visibility and follower count grow further. Thus, enhancing their appeal to other large companies even more than before. This cycle perpetuates a system where the already-popular influencers become even more dominant, often at the expense of smaller, perhaps equally talented creators who struggle to get noticed.

This phenomenon raises concerns about diversity and innovation within the influencer marketing space. As the same set of popular influencers monopolize marketing opportunities, there’s a risk that the content landscape becomes homogenized, with less room for fresh perspectives and unique voices. This stifles creativity and limits the potential for audiences to engage with a broader range of content.

While the allure of working with top influencers is understandable from a marketing perspective, companies must consider the broader implications of their selection strategies. By fostering a more inclusive environment, businesses can drive innovation, engage with a broader audience, and craft meaningful connections in the digital age.

https://www.forbes.com/sites/theyec/2018/07/30/understanding-influencer-marketing-and-why-it-is-so-effective/?sh=1730833171a9

https://www.meltwater.com/en/blog/influencer-fatigue

The Dot Com Bubble and Information Cascade

The late 1990s witnessed the Dot-Com Bubble, a period characterized by an unprecedented surge in investment in internet-related companies. Despite lacking solid business models or clear paths to profitability, these companies achieved extremely high valuations, fueled by venture capitalists and traditional investors who were eager to invest in those companies. However, by the early 2000s, the bubble burst, and the market took a steep dive. 

One of the major reasons why the Dot Com Bubble happened is because of a collective oversight of fundamental valuation principles, but it is also closely related to information cascades that we learned in class. Information cascades happen when individuals start making decisions based on the actions of others and not on their own analysis or data. During the bubble, there were numerous stories of lucrative IPOs, quick wealth from tech stocks, and the growing potential of the Internet sector. This created a sense of FOMO (fear of missing out), compelling individuals to invest based on the momentum, rather than the fundamental evaluation of the companies. 

Historical trends have shown repeatedly how individual investors are influenced by the actions of others and follow the momentum. It’s truly fascinating to see the repeated patterns in the market, which underscores the impact of information cascades. As someone interested in the financial markets and investment strategies, it was very interesting to see how what we learned in class directly applies to real-world scenarios like the Dot-Com Bubble. 

Source: https://www.thestreet.com/dictionary/dot-com-bubble-and-burst

TACO BELLS USE OF SNAPCHAT AS AN EXAMPLE OF A SPONSORED SEARCH

In the dynamic realm of digital marketing, companies are always seeking new and creative methods to engage with their target consumers. Taco Bell, a company known for its innovative digital marketing tactics in addition to its fast-food hunger, is one example of such a breakthrough. Their most recent foray into the realm of Snapchat filters blurs the boundaries between entertainment and advertising and offers an intriguing case study of contemporary marketing techniques. Here’s how Taco Bell’s use of Snapchat filters puts a creative spin on the idea of sponsored search.

The Rise of Snapchat as a Marketing Medium

With its captivating filters and brief content, Snapchat has become an effective tool for marketers trying to connect with younger consumers. Brands hoping to add some humor to their advertising have shown a particular interest in the app’s filters, which let users change their looks or surroundings in creative and amusing ways.

Taco Bell’s Foray into Filtered Fun

The purpose of Taco Bell’s Snapchat filter is to increase brand awareness and interaction in a congested digital world. It’s more than simply a gimmick for the millennial generation. Taco Bell enhances the perception of their brand as lively, youthful, and inventive while expanding its reach by developing a filter that people find entertaining and worthwhile to share.

Sponsored Search, Redefined

In the past, sponsored search has referred to paid advertisements that show up above or next to search results on internet search engines like Google with the intention of drawing users to the advertiser’s website and increasing exposure. Although Taco Bell’s Snapchat filter isn’t a traditional search result, it nonetheless perfectly captures the spirit of promoted search by making sure the brand is always present in a user’s social networking posts.

Visibility Through Virality

Because Snapchat filters are naturally shareable, employing them as a sponsored search option is brilliant. Taco Bell becomes more visible as a consequence of user-generated content that circulates around the network rather than by way of direct search results when users use a filter and share their images or videos with others. Because of its virality impact, the filter functions as an ongoing advertisement that affects everyone in the user’s social circle in addition to the original user.

Engagement Beyond Clicks

This strategy also has the benefit of encouraging a high degree of participation. In contrast to conventional sponsored search advertisements, which depend on consumers going through to a website, a branded Snapchat filter encourages users to communicate directly and personally with the business. Because customers like interacting with the company’s content, this engagement is not only more remembered but also more likely to create a good brand connection.

The Future of Sponsored Search

Taco Bell’s usage of a Snapchat filter as a promotional tool is part of a larger change in the way businesses handle internet advertising. Brands need to come up with innovative strategies to grab customers’ attention and interact with them as they get more skilled at spotting or blocking traditional advertisements. In this regard, innovative applications of social media sites such as Snapchat are expected to be vital to sponsored search in the future.

Conclusion

Taco Bell’s Snapchat filter is a great illustration of how sponsored search is expanding to include social media’s dynamic and interactive potential in addition to search engines. Taco Bell enhances their visibility and creates a lasting impression on users by utilizing the viral popularity of Snapchat filters. Taco Bell’s approach provides insightful information for other businesses seeking to engage their customers in new and meaningful ways as online advertising continues to evolve in response to the shifting internet environment.

Deciphering the Web of Trust: Innovations in Predicting Online Relationships

The article “Predicting Trust and Distrust in Social Networks,” composed by Thomas DuBois, Jennifer Golbeck, and Aravind Srinivasan, presents a run of profitable points of view on the challenge of anticipating trust and distrust in social systems. Most broadcast communications forms nowadays take place within the theoretical environment of the web, and the creators contend that it is best caught on through computational models and strategies. This is often because they work as common speakers, and progressively, reportage rises as an item of the online short-message, user-generated environments, or widespread social systems. In this way, the Web environment could be a mass of information and disinformation. In this environment, signal-to-noise proportion terms, are typically a case where the signal-to-noise especially because it can be assessed and demystified by an extent of computational methods pivotal.

The article presents a modern calculation for recognizing trust and distrust by combining a probabilistic approach to trust utilizing arbitrary charts and an altered spring-embedding calculation. This approach is eminent for its capacity to handle the non-transitive nature of doubt, a critical challenge within the field. Understanding trust and distrust is basic for making strides in client involvement in social stages, counting substance sifting, personalized proposals, and community engagement. Precisely anticipating believe connections have suggestions past the scholarly world, with down-to-earth applications in online suggestion frameworks, substance balance, and the broader objective of cultivating agreeable and dependable online communities.

The researchers’ utilization of computational strategies to address the complicated and multi-dimensional concept of trust may be an essential perspective of their work. By recognizing the confinements of coordinated trust connections and digging into the flow of induced trust and distrust, the ponder altogether contributes to our comprehension of the basic structures of social systems. This investigation not only propels the specialized discussion encompassing trust induction calculations but also emphasizes the significance of making frameworks that account for the unobtrusive complexities of human social networks online. The discoveries of DuBois and colleagues lay the basis for encouraging examination into the instruments of trust, giving roads for creating more advanced and nuanced instruments for exploring the growing computerized social scene.

Source:
https://www.researchgate.net/publication/220876028_Predicting_Trust_and_Distrust_in_Social_Networks
TY – BOOK
AU – Dubois, Thomas
AU – Golbeck, Jennifer
AU – Srinivasan, Aravind
PY – 2011/10/01
SP – 418
EP – 424
T1 – Predicting Trust and Distrust in Social Networks
VL –
DO – 10.1109/PASSAT/SocialCom.2011.56
ER –

How “Infinite Craft” Mirrors Network Theory

I recently came upon an interesting game called “Infinite Craft” on TikTok, where players were sharing their original creations and discoveries. This sandbox browser game, created by Neal Agarwal, begins with the four fundamental elements of earth, wind, fire, and water. Players can then mix and match these elements to construct an endless number of additional elements, which can range from straightforward things to intricate ideas and imaginary creatures. The game’s simplicity, coupled with the vast possibilities for creativity, has made it a viral sensation on social media platforms.

What’s fascinating about “Infinite Craft” is not just its popularity but also how it relates to network theory. In the game, each element can be viewed as a node in a network, with connections formed by the combinations that lead to new elements. As players experiment and discover new combinations, they essentially map out an interconnected network of elements. This mirrors real-world networks, where nodes and connections represent entities and their relationships, whether in social networks, biological systems, or technological frameworks.

Furthermore, “Infinite Craft” provides a playful yet insightful way to explore the dynamics of network growth and evolution. As players add new elements to the network, they can observe how the structure and complexity of the network evolve over time. This can serve as a metaphor for understanding how real-world networks expand and how new connections can lead to emergent properties and behaviors. Overall, “Infinite Craft” is not just a game for entertainment; it can be used as a virtual sandbox for exploring the principles of network theory engagingly and intuitively.

Source: https://neal.fun/infinite-craft/

Instagram Hiding ‘Like’ Counts: Confronting Cascading Behaviors

As social media continues to consume more and more hours of our lives, it often leads to the emergence of trend-based actions and cascading behaviors from its users. In light of artificial personas and social networks’ mental health effects, at the end of April in 2019, Instagram announced its plan to test hiding public ‘like’ counts on videos and photos posted. The user who shares the post on the network would retain visibility of how many people had liked it, while the overall count of likes would remain undisclosed to anyone else. Soon enough, Instagram solidified this update of hiding public displays of like-counts for users globally, as visualized below.

As we learned in Chapter 16 about information cascades, when individuals are connected by a network, they may influence each other’s behaviors and actions. The individuals a user follows on Instagram can be considered to be in a cascade with them, where each person draws rational inferences (decision to also like the post in this case) from information they can see. After the implementation of Instagram’s hidden like-count, the user holds fewer pieces of information relative to the cascade of their network and they are hence less inclined to continue the feedback loop by liking the picture. The article chosen illustrates how Instagram hoped that by hiding their users’ posts’ public like-counts, users would resist cascade effects by not being able to follow the crowd as closely as before. Instead, users may increasingly like posts they truly appreciate without facing public consequences. Cascade effects are also anticipated to decrease as influencers will feel less pressure to compare themselves to others, or to post content just because they believe it will perform well. 

Prior to the implementation of like-count hiding and optional comment-count hiding, a cascade could begin when an individual sees that multiple of their friends on Instagram have also liked a certain post and it’s liked by a large number of other people. The posts with higher total likes would continue to be engaged with, making popular posts even more popular. In fact, Beca Alexander from this 2019 article explains “since likes are currently a way to measure the popularity of any given post, removing the quantitative value placed on each piece of content will relieve users of the visibility of what their ‘like’ action choice will or will not do for each piece of content.” In the context of the power law distribution, popularity is characterized by extreme imbalances, with very large popularity values arising from certain posts. 

In theory, when the like-counts were public, we’d anticipate that posts’ popularity distribution were even more skewed due to rich-get-richer effects. The probability that a post experiences an increase in popularity is directly proportional to its current popularity through its like-count and engagement, which is still true, but to a lower extent with the like-count being hidden. This closely mirrors the course’s examples of how publishing a site’s viewer count publicly adjacent to it led to the emergence of the power law and a phenomenon where the already popular content tends to attract more attention, thus creating a rich-get-richer effect.

Chapter 12’s illustration below of how a node’s position in a network affects its power in the market also comes up in this article. The article explains that Instagram influencers who currently hold excellent engagement and use it as a bargaining chip will have less power in brand negotiations after the hidden like-count. Their ability to garner social media popularity may have translated into increased social capital within their circle, granting them significant influence and leverage in group dynamics. While that influencer with higher likes may have been the most powerful node in their social network and brand, their leverage and bargaining power may be diminished without public display of likes.

Instagram’s decision to hide like counts on posts in 2021 marked a significant shift in the platform’s dynamics, aimed at mitigating the negative impacts of cascading behaviors and enhancing user experience towards authenticity. However, post-implementation in 2023 research revealed mixed responses, suggesting that the change had varying effects on different users. While some users appreciated the shift away from public like counts, others found it inconvenient, as they relied on such metrics to gauge trends and popularity. Instagram’s acknowledgment of this diversity in user preferences led to the implementation of a choice-based approach, allowing users to toggle the visibility of like counts according to their preferences. Overall, Instagram realized that cascading effects and public like-count information could be beneficial as they learned that many of their users value viewing which of their followers and friends have also liked a post, hence influencing their own decisions.

Article Link: https://hellopartner.com/2019/05/14/what-is-potential-impact-of-instagram-hiding-like-counts/

Subways, Trains, and Rail – Americas Underserved and Underutilized Network

Commentary on this article: https://www.theguardian.com/books/2023/nov/14/book-lost-subways-north-america-jake-berman#:~:text=If%20there%20are%20common%20factors,to%20make%20said%20routes%20thrive.

America has notoriously bad public transit networks, granted a few exceptions. Public transit is essential to most developed cities across the world, however this is uniquely not the case in America. Multiple major American cities are virtually unlivable without consistent access to a personal motor vehicle. While arguably this is partly due to cultural elements in America that lead to car dependency such as this nation’s tendency towards rampant individualism, but it is also due to how poorly designed many public transit networks are in the USA.

The networks are often designed in ways that make them inefficient, expensive, and difficult to use for riders. Many of the systems have nodes that are not particularly close to major centers of activity meaning they aren’t giving riders direct access to the places they are most likely to want to go. An important philosophy espoused regarding public transportation is more stations (nodes) does not necessarily equal more accessibility or utility, if these nodes are not placed optimally. In fact, poorly placed nodes can actually harm the system by slowing it down and making it more complex without much added benefit. Public Transport grids are a great example of a real world network with nodes, connection points, etc. These networks do not function well in America due to poorly placed nodes, inefficient connection points, and missing connection points.