When would online platforms pay data dividends Sukanya Kudva and Anil Aswani Abstract Online platforms including social media and

2025-05-01 0 0 285.92KB 6 页 10玖币

侵权投诉

When would online platforms pay data dividends?

Sukanya Kudva and Anil Aswani

Abstract— Online platforms, including social media and

search platforms, have routinely used their users’ data for

targeted ads, to improve their services, and to sell to third-party

buyers. But an increasing awareness of the importance of users’

data privacy has led to new laws that regulate data-sharing

by platforms. Further, there have been political discussions on

introducing data dividends, that is paying users for their data.

Three interesting questions are then: When would these online

platforms be incentivized to pay data dividends? How does

their decision depend on whether users value their privacy

more than the platform’s free services? And should platforms

invest in protecting users’ data? This paper considers various

factors affecting the users’ and platform’s decisions through

utility functions. We construct a principal-agent model using

a Stackelberg game to calculate their optimal decisions and

qualitatively discuss the implications. Our results could inform

a policymaker trying to understand the consequences of man-

dating data dividends.

I. INTRODUCTION

The revenue models of many online platforms depend on

collecting, analyzing, and selling users’ data. A free-service

and advertising-based revenue model can cause conﬂicts in

the users’ and platform’s interests. Users may be concerned

about their privacy and possible misuse of data, while

platforms want to maximize their proﬁts. Further, users’

perception of a platform’s ethics and their willingness to

participate can be affected by the platform’s revenue model,

pricing decisions, and privacy practices [1]–[3].

A. Cybersecurity on online platforms

After the onset of the Internet of Things (IoT), there

has been an increase in the variety, speed, and volume

of users’ data collected. Information from multiple sources

including devices, sensors, and social networks is being used

by platforms to assist users and collect data [4].

Though users have become more aware of the rampant

collection of their data, they often do not know about

the proliferation of IoT devices in their everyday lives.

For instance, in August 2022 the Australian federal court

convicted a major search platform for collecting users’

location data without their knowledge [5]. When users give

consent and permissions to apps and platforms, they tend to

underestimate the implications of it [6]. Reforms to protect

users’ privacy are very much needed across the world [7].

Users should be able to control, delete and transfer their data

across different platforms and service providers. They should

*This material is based upon work supported by the National Science

Foundation under Grant CMMI-1847666.

S Kudva and A Aswani are with the Department of Indus-

trial Engineering and Operations Research, University of California,

Berkeley, CA 94720, USA sukanya kudva@berkeley.edu,

aaswani@berkeley.edu

be asked for explicit consent every time a platform wants to

use their data for a new purpose [4].

B. Privacy and data dividends

With growing user concerns, consumer privacy legislation

has become an important topic for public discussion, and

multiple new data privacy laws have been introduced [8]. In

Europe, the General Data Protection Regulation (GDPR) was

introduced in 2016 to give consumers more control over their

data [9]. In 2019, legislators in California announced their

intent to introduce data dividends, which is a model in which

platforms would pay users in exchange for use of users’ data

[10]. The same year, Oregon legislators introduced a bill,

called the Health Information Property Act, to compensate

consumers for monetizing their health data [11].

Implementing data dividends comes with its own chal-

lenges as companies holding the data are far more powerful

than individual users. Further, there is a huge information

asymmetry and only companies know the actual value of the

users’ data. The critics of data dividends argue that selling

data would make it a commodity and be counter-productive

in protecting users’ privacy. They also feel that vulnerable

groups – such as people of color and the poor – who are

currently discriminated against should not be incentivized

to pour more data into the system and further reduce their

privacy [12], [13]. On the other hand, the proponents of data

dividends argue that today’s technology economy is hugely

driven by monetizing users’ data, and paying users a share

of these beneﬁts is only fair.

Recent studies have explored different ways of pricing

data dividends for each user based on the value of their

individual data [14], [15]. Using the idea of Shapley and

Owen values, they calculated the contribution made by each

user to the platform’s proﬁts. Some scholars have also

proposed different ways of charging data dividends and how

they could be used for the greater public good [16]. For our

work, we ask different questions: Should platforms pay data

dividends at all, and why? In this paper, we do our analysis

with homogeneous users but our methods can be extended

to analyze heterogeneous users too.

C. Contributions and outline

Our paper is organized as follows: Sect. II outlines our

utility functions for an online platform and its users and

our principal-agent model, Sect. III analytically solves the

principal-agent model in order to derive their optimal choices

and Sect. IV discusses insights from our model.

Our paper comes in the context of rising debates on data

dividends. We try to understand when and how much online

arXiv:2210.01900v1 [cs.GT] 4 Oct 2022

platforms would pay as data dividends. We consider users’

privacy concerns as an important factor, for which a platform

can invest in data protection. The platform can also pay users

to incentivize taking risks and sharing their data. Our main

contributions are to introduce a principal-agent model using

a Stackelberg game to capture the platform-users dynamics

and then use this to gain a better understanding of when the

platform would pay users with data dividends.

II. OUR MODEL

An online platform provides its users with free services

and a data dividend in exchange for their data. The users

are allowed to choose between two levels – high or low

– of data sharing. For each level, the platform provides a

different data dividend and set of services. As the platform

collects users’ data, it faces the risk of a possible data breach.

If a data breach occurs, the platform loses its reputation

and faces possible legal and ﬁnancial complications, and the

users are harmed by the misuse of their data. Hence, the

platforms consider investing in protecting their users’ data.

This could include the costs of building better technological

infrastructure and signing insurance contracts.

A. Platform’s utility

The platform has a ﬁxed cost of Sto maintain and provide

its services. It invests Iin users’ data protection, reducing

the probability of a data breach to B(I). Here, Bis assumed

to be a twice differentiable function such that B(I)>0

with limI→∞ B(I)=0,B0(I)<0and B00(I)>0∀I. In

case of a data breach, the platform has a loss of Ffrom

legal cases and a lost reputation. When having access to k

users’ data, the platform makes a revenue of U(k, b), where

bof the total kusers chose the low level of data sharing.

The revenue may be due to selling the data to a third-party,

selling advertisements to display to users or other sources.

The platform pays a share of this revenue to users as data

dividends, which are priced at two scales – p0and p1– for

the low and high levels of data sharing respectively.

Considering these costs, revenue, and risks, the platform

has a total expected utility of :

U(k, b)−B(I)F−I−S−p0b−p1(k−b).(1)

B. User’s utility

We consider khomogeneous users who have the same

behavior and parameters. Each user values their personal data

at Vand faces an additional personal loss of Lon a data

breach. They also beneﬁt from the platform’s free services,

which amount to a value of W. When a user chooses the low

level of data sharing, these beneﬁts and losses scale down

by a factor α∈(0,1). The platform pays users with data

dividends at two scales of pay and thereby encourages a

particular level of data sharing.

Let cibe user i’s decision variable so that ci= 0 and 1

for the low and high levels of data sharing respectively. Then

a user i’s total expected utility is:

ci(p1− V) + (1 −ci)(p0−αV),(2)

where V=¯

V+B(I)Land ¯

V=V−W.

C. Principal-agent formulation

We construct a Stackelberg game [17]–[19], in which

players act sequentially with follower(s) acting after a leader.

In our model, the platform ﬁrst prices data dividends for

different levels of data-sharing, and decides on investment

for users’ data protection. Given this information, the users

decide how much data to share.

The optimal, equilibrium choices of the platform and the

users are their Stackelberg strategies. We capture this using

an optimization problem, which maximizes the platform’s

utility when the users are maximizing their utilities [20]. It

is formulated as follows:

max U(k, b)−B(I)F−I−S−p0b−p1(k−b)

s.t c∗

i=arg max

ci∈{0,1}ci(p1− V) + (1 −ci)(p0−αV)

∀i={1,· · · , k}

c∗

i(p1− V) + (1 −c∗

i)(p0−αV)≥0

∀i={1,· · · , k}

I, p0, p1≥0.

(3)

The second constraint in (3) ensures that the users do not

have a net loss from using the platform, without which they

would leave the platform.

III. OPTIMAL CHOICES

Since the users are homogeneous, they make similar

choices: either c∗

i= 1 or c∗

i= 0 for every user i. If users

ﬁnd both levels of data sharing to be utility-maximizing, then

we assume that they all choose the level that most beneﬁts

the platform. We solve these two cases of c∗

iseparately. The

optimal solution is the best of the two cases and can vary

with the numerical values of the parameters of the model.

A. Case 1: c∗

i= 1 ∀i={1,· · · , k}

To ensure users choose high level of data sharing, the

platform must price the data dividends such that p1− V ≥

p0−αV. Note that b= 0 in this case. The optimization

problem then reduces to:

max U(k, 0) −B(I)F−I−p1k−S

s.t p1− V ≥ p0−αV

p1− V ≥ 0

I, p0, p1≥0.

(4)

Since p0≥0,p∗

0= 0 is optimal for the problem. Given

this, one can conclude p∗

1=Vwhen V ≥ 0and p∗

1= 0

when V ≤ 0. Also, U(k, 0) −Scan be treated as a constant.

These observations further reduce (4) to two sub-cases with

optimization problems in a single variable I.

1) Sub-case 1 - If V ≥ 0then p∗

1=V:Set p∗

1=V=

V+B(I)Land add an additional constraint V ≥ 0in (4).

min B(I)F1+I+¯

V k

s.t ¯

V+B(I)L≥0

I≥0,

(5)

where F1=F+Lk.

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

Whenwouldonlineplatformspaydatadividends?SukanyaKudvaandAnilAswaniAbstractOnlineplatforms,includingsocialmediaandsearchplatforms,haveroutinelyusedtheirusers'datafortargetedads,toimprovetheirservices,andtoselltothird-partybuyers.Butanincreasingawarenessoftheimportanceofusers'dataprivacyhasledtonewla...

展开>> 收起<<

When would online platforms pay data dividends Sukanya Kudva and Anil Aswani Abstract Online platforms including social media and.pdf

共6页,预览2页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

When would online platforms pay data dividends Sukanya Kudva and Anil Aswani Abstract Online platforms including social media and

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: