About
I am a data science and machine learning leader specializing at helping companies build…
Articles by Fangfang
-
NABE TECH 2019 Conference Registration Calls for Early-Round Registration
NABE TECH 2019 Conference Registration Calls for Early-Round Registration
By Fangfang Tan
Activity
-
3 hard truths about careers in machine learning: 1️⃣ You'll realize that 90% of AI in the media is BS With the rise of “AI”, companies are abusing…
3 hard truths about careers in machine learning: 1️⃣ You'll realize that 90% of AI in the media is BS With the rise of “AI”, companies are abusing…
Liked by Fangfang Tan
-
In my most recent talks to companies, even though everyone is talking about AI, less than 10% of people (these are mostly senior managers) have even…
In my most recent talks to companies, even though everyone is talking about AI, less than 10% of people (these are mostly senior managers) have even…
Liked by Fangfang Tan
-
It's a wrap! FY24 was an intense year and our Connected GTM Ops team continued to deliver exceptional impact for LinkedIn. With our transition to…
It's a wrap! FY24 was an intense year and our Connected GTM Ops team continued to deliver exceptional impact for LinkedIn. With our transition to…
Liked by Fangfang Tan
Experience
Education
Licenses & Certifications
-
NCLC Attendence 2024
Asia Society
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Volunteer Experience
-
Board Member
The Joy Culture Foundation
- Present 1 year
Education
Board member and part of the Operation & Program Committee. Obligations include:
- To oversee new program development, and to monitor and assess existing programs;
- To initiate and guide program evaluations;
- To facilitate discussions about program priorities for the operations;
- Develop best practices for operation models
Publications
-
Third-party punishment: Retribution or deterrence?
Journal of Economic Psychology
-
Providing global public goods: electoral delegation and cooperation
Economic Inquiry
This study experimentally examines the effect of electoral delegation on providing global public goods shared by several groups. Each group elects one delegate who can freely decide on each group member's contribution to the global public goods. Our results show that people mostly vote for delegates who assign equal contributions for every group member. However, in contrast to standard theoretical predictions for our delegation mechanism, unequal contributions across groups drive cooperation…
This study experimentally examines the effect of electoral delegation on providing global public goods shared by several groups. Each group elects one delegate who can freely decide on each group member's contribution to the global public goods. Our results show that people mostly vote for delegates who assign equal contributions for every group member. However, in contrast to standard theoretical predictions for our delegation mechanism, unequal contributions across groups drive cooperation down over time, and it decreases efficiency by almost 50% compared to the selfish benchmark. This pattern is not driven by delegates trying to exploit their fellow group members, as indicated by theory. It is driven by conditional cooperation of delegates across groups. Since one of the potential sources of the resulting inefficiency is the polycentric nature of global public goods provision together with other-regarding preferences, we use the term P-inefficiency to describe our finding.
Other authors -
-
How could economists find their roles in tech companies (in Chinese)
TalkEcon
This essay aims at sharing some of my thoughts transitioning from academia to industry. Concretely, it summarizes some common types of roles economists are currently taking at tech companies, outlines some mindset challenges during the transition and suggestions to overcome them.
-
'Success Breeds Success' or 'Pride Goes Before a Fall'? Teams and Individuals in Multi-contest Tournaments
Games and Economic Behavior
We study the impact of progress feedback on players' performance in multi-battle team contests, in which team members' efforts are not directly substitutable. In particular, we employ a real-effort laboratory experiment to understand, in a best-of-three-contest setting, how players' strategic mindsets change when they compete on a team compared to when they compete individually. Our data corroborate the theoretical predictions for teams: Neither a lead nor a lag in the first component…
We study the impact of progress feedback on players' performance in multi-battle team contests, in which team members' efforts are not directly substitutable. In particular, we employ a real-effort laboratory experiment to understand, in a best-of-three-contest setting, how players' strategic mindsets change when they compete on a team compared to when they compete individually. Our data corroborate the theoretical predictions for teams: Neither a lead nor a lag in the first component contest affects a team's performance in the subsequent contests. In individual tournaments, however, contrary to the theoretical prediction, we observe that leaders perform worse - but laggards perform better - after learning the outcome of the first battle. Our findings offer the first empirical evidence from a controlled laboratory on the impact of progress feedback between team and individual contests, and contribute new insight on team incentives.
Other authorsSee publication -
E-commerce Recommendation with Personalized Promotion
RecSys '15 Proceedings of the 9th ACM Conference on Recommender Systems
Most existing e-commerce recommender systems aim to recommend the right products to a consumer, assuming the properties of each product are fixed. However, some properties, including price discount, can be personalized to respond to each consumer's preference. This paper studies how to automatically set the price discount when recommending a product, in light of the fact that the price will often alter a consumer's purchase decision. The key to optimizing the discount is to predict consumer's…
Most existing e-commerce recommender systems aim to recommend the right products to a consumer, assuming the properties of each product are fixed. However, some properties, including price discount, can be personalized to respond to each consumer's preference. This paper studies how to automatically set the price discount when recommending a product, in light of the fact that the price will often alter a consumer's purchase decision. The key to optimizing the discount is to predict consumer's willingness-to-pay (WTP), namely, the highest price a consumer is willing to pay for a product. Purchase data used by traditional e-commerce recommender systems provide points below or above the decision boundary. In this paper we collected training data to better predict the decision boundary. We implement a new e-commerce mechanism adapted from laboratory lottery and auction experiments that elicit a rational customer's exact WTP for a small subset of products, and use a machine learning algorithm to predict the customer's WTP for other products. The mechanism is implemented on our own e-commerce website that leverages Amazon's data and subjects recruited via Mechanical Turk. The experimental results suggest that this approach can help predict WTP, and boost consumer satisfaction as well as seller profit.
Other authorsSee publication -
Justification and legitimate punishment
Journal of Institutional and Theoretical Economics
Punishment can lose its legitimacy if the enforcer can profit from delivering punishment. We examine how justification can promote the legitimacy of punishment in a one-shot sender–receiver game where an independent third party can punish the sender upon seeing whether the sender lied. Most third parties who can profit from punishment punish the senders regardless of how the senders behave. However, when they have to provide explanations for their punishment decisions, significantly more third…
Punishment can lose its legitimacy if the enforcer can profit from delivering punishment. We examine how justification can promote the legitimacy of punishment in a one-shot sender–receiver game where an independent third party can punish the sender upon seeing whether the sender lied. Most third parties who can profit from punishment punish the senders regardless of how the senders behave. However, when they have to provide explanations for their punishment decisions, significantly more third parties punish the sender if and only if the sender lies, and senders are also more likely to perceive punishment as legitimate and behave honestly.
Other authors -
-
Can strategic uncertainty help deter tax evasion? - An experiment on auditing rules
Journal of Economic Psychology
This paper adds to the economic-psychological research on tax compliance by experimentally testing a simple auditing rule that induces strategic uncertainty among taxpayers. Under this rule, termed the bounded rule, taxpayers are informed of the maximum number of audits by a tax authority, so that the audit probability depends on the joint decisions among the taxpayers. We compare the bounded rule to the widely studied flat-rate rule, where taxpayers are informed that they will be audited with…
This paper adds to the economic-psychological research on tax compliance by experimentally testing a simple auditing rule that induces strategic uncertainty among taxpayers. Under this rule, termed the bounded rule, taxpayers are informed of the maximum number of audits by a tax authority, so that the audit probability depends on the joint decisions among the taxpayers. We compare the bounded rule to the widely studied flat-rate rule, where taxpayers are informed that they will be audited with a constant probability. The experimental evidence shows that, as theoretically predicted, the bounded rule induces the same level of compliance as the flat-rate rule when strategic uncertainty is low, and a higher level of compliance when strategic uncertainty is high. The bounded rule also induces distinctive tax evasion dynamics compared to the flat-rate rule. The results suggest that increasing the level of strategic uncertainty among taxpayers could be an effective device to deter tax evasion.
► We examine a bounded rule inducing strategy uncertainty among taxpayers.
► We experimentally compare the bounded rule with a flat-rate audit rule.
► Under the bounded rule, participants are informed of the maximum audit number.
► The bounded rule deters cheaters more effectively when strategic uncertainty is high.
► The evasion dynamics of the bounded rule differs from that of the flat-rate rule.Other authors -
-
Who acts more like a game theorist? Group and individual play in a sequential market game and the effect of the time horizon
Games and Economic Behavior
Previous experimental results on one-shot sequential two-player games show that group decisions are closer to the subgame-perfect Nash equilibrium than individual decisions. We extend the analysis of intergroup versus interindividual decision-making by running both one-shot and repeated sessions of a simple two-player sequential market game (Stackelberg duopoly). Whereas in one-shot markets we find no significant differences in the behavior of groups and individuals, in repeated markets we find…
Previous experimental results on one-shot sequential two-player games show that group decisions are closer to the subgame-perfect Nash equilibrium than individual decisions. We extend the analysis of intergroup versus interindividual decision-making by running both one-shot and repeated sessions of a simple two-player sequential market game (Stackelberg duopoly). Whereas in one-shot markets we find no significant differences in the behavior of groups and individuals, in repeated markets we find that the behavior of groups is further away from the subgame-perfect equilibrium of the stage game than that of individuals. To a large extent, this result is independent of the method of eliciting choices (sequential or strategy method), the matching protocol (random- or fixed-matching), and the econometric method used to account for observed first- and second-mover behavior. We discuss various possible explanations for the differential effect that the time horizon of interaction has on the extent of individual and group playersʼ (non)conformity with subgame perfectness.
Other authors -
-
Peer punishment with third-party approval in a social dilemma game
Economics Letters
In a prisoner’s dilemma experiment, compared with the case when the implicated parties are allowed to punish each other, both the cooperation rate and the earnings are lower when the enforcement of punishment requires approval from an independent third party.
► Sanction requests often need approvals from an independent third party.
► Cooperation rate and average earnings are lower when a third party can veto sanctions.
► The intervention of independent third parties reduces the…In a prisoner’s dilemma experiment, compared with the case when the implicated parties are allowed to punish each other, both the cooperation rate and the earnings are lower when the enforcement of punishment requires approval from an independent third party.
► Sanction requests often need approvals from an independent third party.
► Cooperation rate and average earnings are lower when a third party can veto sanctions.
► The intervention of independent third parties reduces the severity of punishment.Other authors -
-
Voting on punishment systems with a heterogeneous group
Journal of Public Economic Theory
We consider a voluntary contributions game, in which players may punish others after contributions are made and observed. The productivity of contributions, as captured in the marginal-per-capita return, differs among individuals, so that there are two types: high and low productivity. Every two or eight periods, depending on the treatment, individuals vote on a punishment regime, in which certain individuals are permitted, but not required, to have punishment directed toward them. The…
We consider a voluntary contributions game, in which players may punish others after contributions are made and observed. The productivity of contributions, as captured in the marginal-per-capita return, differs among individuals, so that there are two types: high and low productivity. Every two or eight periods, depending on the treatment, individuals vote on a punishment regime, in which certain individuals are permitted, but not required, to have punishment directed toward them. The punishment system can condition on type and contribution history. The results indicate that the most effective regime, in terms of contributions and earnings, is one that allows punishment of low contributors only, regardless of productivity. Nevertheless, only a minority of sessions converge to this system, indicating a tendency for the voting process to lead to suboptimal institutional choice.
Other authors -
-
Punishment in a linear public good game with productivity heterogeneit
De Economist
Although it is widely recognized that sanction increases cooperation in a public good game, comparatively little attention has been paid to a scenario in which agents have heterogeneous productivities (i.e. asymmetric impact on the group account). This paper examines the extent to which sanction works in this scenario by varying marginal per capita return (MPCR) among group members. Experimental results indicate that in the absence of sanctions, productivity heterogeneity hampers cooperation…
Although it is widely recognized that sanction increases cooperation in a public good game, comparatively little attention has been paid to a scenario in which agents have heterogeneous productivities (i.e. asymmetric impact on the group account). This paper examines the extent to which sanction works in this scenario by varying marginal per capita return (MPCR) among group members. Experimental results indicate that in the absence of sanctions, productivity heterogeneity hampers cooperation. Allowing punishment in these groups significantly enhances average contributions of group members, but does not increase welfare. In groups in which cooperation is highly successful, high-productivity agents actively punish low-productivity agents in initial periods. However, conditional on individual contributions, high-productivity agents receive more punishment, and behave more responsively by raising their contributions in the next period. The results mirror the reality in which elites in a society are under higher pressure, since their choices are likely to have a deeper impact on a society.
Patents
-
Label Shift Detection and Adjustment in Predictive Modeling
Issued 11599746
Techniques for detecting label shift and adjusting training data of predictive models in response are provided. In an embodiment, a first machine-learned model is used to generate a predicted label for each of multiple scoring instances. The first machine-learned model is trained using one or more machine learning techniques based on a plurality of training instances, each of which includes an observed label. In response to detecting a shift in observed labels, for each segment of one or more…
Techniques for detecting label shift and adjusting training data of predictive models in response are provided. In an embodiment, a first machine-learned model is used to generate a predicted label for each of multiple scoring instances. The first machine-learned model is trained using one or more machine learning techniques based on a plurality of training instances, each of which includes an observed label. In response to detecting a shift in observed labels, for each segment of one or more segments in multiple segments, a portion of training data that corresponds to the segment is identified. For each training instance in a subset of the portion of training data, the training instance is adjusted. The adjusted training instance is added to a final set of training data. The machine learning technique(s) are used to train a second machine-learned model based on the final set of training data.
Other inventorsSee patent
Courses
-
Advanced Game Theory
-
-
Behavioral Economics
-
-
Economic Foundations of Organization, Strategy and International Business
-
-
Economics of Psychology
-
-
Experimental Economics
-
-
Microeconometrics
-
-
Microeconomics
-
-
Panel Data Analysis
-
-
Time Series Econometrics
-
Projects
-
Optimizing Audience Buying on Facebook and Instagram
Using randomized controlled trails (CRT), this paper shows that placement optimization is a cost-effective way for brand advertisers to extend their Facebook campaigns to other platforms like Instagram and Audience Network.
Other creatorsSee project -
Drive Action Effectively on Mobile: Optimize Direct Response Campaigns across Facebook, Instagram and Audience Network.
-
I used randomized control experiments (RCT) to measure the value in using placement optimization (PO) for DR objectives. The results showed that DR campaigns which used PO resulted in greater reach, more incremental converters and more incremental conversions than delivering to Facebook alone.
TOPLINE RESULTS:
- 8 of the 11 campaigns were found to have statistically significant incremental lift when optimized across Facebook platforms.
- Reach in the placement optimized campaigns…I used randomized control experiments (RCT) to measure the value in using placement optimization (PO) for DR objectives. The results showed that DR campaigns which used PO resulted in greater reach, more incremental converters and more incremental conversions than delivering to Facebook alone.
TOPLINE RESULTS:
- 8 of the 11 campaigns were found to have statistically significant incremental lift when optimized across Facebook platforms.
- Reach in the placement optimized campaigns were as much as 7% higher than Facebook-alone and generated 1.67x incremental converters with 3.45x incremental conversions. -
Measurement for Success
-
This paper illustrates how Facebook measures its own brand campaign effectiveness on the platform of Facebook and Instagram.We pick three main questions often asked by digital brand marketers. The key message is to Identify the most important business objectives for your brand and prioritize the right research questions to evaluate your campaign’s effectiveness.
Other creatorsSee project
Honors & Awards
-
CREED Small Grant for Experimental Projects
University of Amsterdam
-
Excellent Teacher Award
School of Economics and Management
-
HSP Huygens Scholarship
Nuffic (Netherlands organisation for international cooperation in higher education)
-
CentER Koopmans Scholarship
Tilburg University
-
Excellent Degree Paper Award
School of Economics, Zhejiang University
Languages
-
English
Full professional proficiency
-
Chinese (Cantonese)
Native or bilingual proficiency
-
Chinese (Mandarin)
Native or bilingual proficiency
Recommendations received
8 people have recommended Fangfang
Join now to viewMore activity by Fangfang
-
Meet Vivian, our Program Director at The Joy Culture Foundation! Vivian Huang is a seasoned educator with 18 years of experience teaching Chinese to…
Meet Vivian, our Program Director at The Joy Culture Foundation! Vivian Huang is a seasoned educator with 18 years of experience teaching Chinese to…
Liked by Fangfang Tan
-
𝐓𝐡𝐞 𝐕𝐞𝐧𝐭𝐮𝐫𝐞 𝐌𝐢𝐧𝐝𝐬𝐞𝐭 𝐢𝐧 𝐒𝐅𝐎 😊 On my way to leading a strategic offsite session at a global VC firm, I made a quick stop at…
𝐓𝐡𝐞 𝐕𝐞𝐧𝐭𝐮𝐫𝐞 𝐌𝐢𝐧𝐝𝐬𝐞𝐭 𝐢𝐧 𝐒𝐅𝐎 😊 On my way to leading a strategic offsite session at a global VC firm, I made a quick stop at…
Liked by Fangfang Tan
-
Celebrating DoorDash’s 11th Anniversary with Tony Xu During my 4+ years at DoorDash, I've learned a lot. Here are my top takeaways: - Bias for…
Celebrating DoorDash’s 11th Anniversary with Tony Xu During my 4+ years at DoorDash, I've learned a lot. Here are my top takeaways: - Bias for…
Liked by Fangfang Tan
-
July marks the start of a new fiscal year for LinkedIn, and I’m excited to welcome the marketing team and the Economic Graph Research Institute, led…
July marks the start of a new fiscal year for LinkedIn, and I’m excited to welcome the marketing team and the Economic Graph Research Institute, led…
Liked by Fangfang Tan
-
If you’ve come to machine learning through data science, this book by Alice Zheng and amanda casari does an amazing job at filling in any remaining…
If you’ve come to machine learning through data science, this book by Alice Zheng and amanda casari does an amazing job at filling in any remaining…
Liked by Fangfang Tan
-
Looks like my paper "Tabular Data: Deep Learning is Not All You Need" just hit 1,000+ citations 😳 Here's the story of how we almost didn't…
Looks like my paper "Tabular Data: Deep Learning is Not All You Need" just hit 1,000+ citations 😳 Here's the story of how we almost didn't…
Liked by Fangfang Tan
-
After an incredible 11-year journey with the Siri team, I am retiring from Apple. While I will miss the remarkable team at Apple, I am excited to…
After an incredible 11-year journey with the Siri team, I am retiring from Apple. While I will miss the remarkable team at Apple, I am excited to…
Liked by Fangfang Tan
-
Many friends, colleagues and parents asked me this question: What could kids learn from using AI apart from enhancing the traditional skills in…
Many friends, colleagues and parents asked me this question: What could kids learn from using AI apart from enhancing the traditional skills in…
Shared by Fangfang Tan
Other similar profiles
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore MoreOthers named Fangfang Tan
-
Fangfang Tan
Purchase Coordinator at Hot Mix Oy Finland
-
fangfang tan
Founder/first executive officer of Liqian Hardware Company, specializing in manufacturing and sales as a whole, providing solutions for hardware tools, surviving on quality and winning reputation with integrity.
-
Fangfang Tan
Junior Cloud Developer bei p36 GmbH
-
Fangfang Tan
R&D Assistant
76 others named Fangfang Tan are on LinkedIn
See others named Fangfang Tan