Building a healthier feed Private location trace intersection driven feed recommendations_2

2025-04-27 0 0 955.98KB 8 页 10玖币

侵权投诉

BUILDING A HEALTHIER FEED:

PRIVATE LOCATION TRACE INTERSECTION DRIVEN

FEED RECOMMENDATIONS

A PREPRINT

Tobin South1,2,*, Nick Lothian3, Takahiro Yabe1, and Alex ‘Sandy’ Pentland1,2

1MIT Connection Science, Massachusetts Institute of Technology, Cambridge, USA

2MIT Media Lab, Massachusetts Institute of Technology, Cambridge, USA

3Verida, Adelaide, Australia

*tsouth@mit.edu

September 21, 2023

ABSTRACT

The physical environment you navigate strongly determines which communities

and people matter most to individuals. These effects drive both personal access to

opportunities and the social capital of communities, and can often be observed in the

personal mobility traces of individuals. Traditional social media feeds underutilize

these mobility-based features, or do so in a privacy exploitative manner. Here we

propose a consent-ﬁrst private information sharing paradigm for driving social

feeds from users’ personal private data, speciﬁcally using mobility traces. This

approach designs the feed to explicitly optimize for integrating the user into the

local community and for social capital building through leveraging mobility trace

overlaps as a proxy for existing or potential real-world social connections, creating

proportionality between whom a user sees in their feed, and whom the user is likely

to see in person. These claims are validated against existing social-mobility data,

and a reference implementation of the proposed algorithm is built for demonstration.

In total, this work presents a novel technique for designing feeds that represent real

ofﬂine social connections through private set intersections requiring no third party,

or public data exposure.

Keywords Social Media Feeds

Private Data Sharing

Personal Data Stores

Mobility Data

Note: This work has been published in the proceedings of SBP-BRiMS 2023 and is posted on

arXiv for ease of access. The manuscript can be accessed at Social, Cultural, and Behavioral

Modeling. Lecture Notes in Computer Science, vol 14161. Springer, Cham. 16 September 2023.

https://doi.org/10.1007/978-3-031-43129-6_6

1 Introduction

Modern social media platforms design feeds to explicitly optimize for user attention, creating negative

effects for both users and broader democratic discourse [

]. These feeds create a feedback loop of

preferential attachment to already popular content and users, leading to disproportionally dominant

agents [

]. To mitigate this rich-get-richer regime, feeds need to focus on promoting low-popularity

content as well, but this is not sufﬁcient — you can’t just show people low-quality content or users

stop engaging. Rather than promoting the content of the largest creators or random content pulled

from across the social internet, we want to optimize social capital [3].

arXiv:2210.01927v2 [cs.CY] 20 Sep 2023

Private location trace intersection feed

Ideally, we want platforms that achieve sustainability and longer-term value by increasing real-world

community connections, trust, capacity for collective action (known as bonding social capital), as

well the number of trusted links to people in other communities (known as bridging capital). At its

core, your feed should be representative of people that matter to you.

The physical spaces we navigate are strong determinants of many aspects of our lives and values,

from our wellbeing [

] to our choice of collaborators [

] and the opportunities for economic growth

we’re exposed to [

]. Location history is a better predictor of your interests than the demographics

used by current matching apps (e.g., dating app, content recommendation, etc.) [

], and is strongly

predictive of which friendships you already have [8]. Leveraging GPS location data in social media

has a long history demonstrating its value, but has been a series of privacy nightmares that require the

sharing of this location data to a third party, often to be stored indeﬁnitely and monetized without any

visibility or accountability to the user.

We propose a new approach to building social media feeds that are optimized for social capital

building via private matching over location histories, where friends are shown in proportion to their

potential to build new within-community relationships and to reinforce community social capital,

requiring no third-party interactions or capture of data.

All data sharing is done securely such that no parties ever see anyone else’s location data, and only

each pair of users can see the size of the location data overlap between them.

This approach is naturally extensible to a huge range of similar matching approaches ranging from

shared photo matching, or shared friendship matching (including from existing social networks), to

simple interest matching, all without needing trusted third parties or the exposure of any personal

information. In doing so, this creates an in-built incentive for users to make personal data accessible

(in an encrypted and private manner) to ensure they appear on others’ feeds and can participate in the

social ecosystem.

In summary, this paper:

•

Presents a novel approach to building a social feed that represents real social relationships

via calculations of set intersections between pairs of individuals without the need for a third

party or transmission of any unencrypted data, ensuring no escape of any private information

to anyone.

• Validates the value of location trace based matching as a measure of friendship on existing

data using two different computational examinations, in turn motivating its use in building a

feed optimized for social capital building.

•

Prototypes the approach with a reference implementation and demo app to show that this

can be done with extremely low computational cost.

•

Proposes and discusses the extensibility of this private set intersection driven feed paradigm

to other personal datasets and explores how this creates new social incentives around private

data sharing for matching.

A longer, more verbose version of this publication with additional ﬁgures will be made available

online via pre-print servers due to page limits.

2 A social recommendation algorithm

For two friends Alice & Bob, the rate at which content is shown on a feed should be proportional to

the physical-world relevancy of their ties (here the likelihood they’ll be at the same location).

When Bob friends Alice, Bob can make a request to Alice to begin a multistep private set intersection

on the mobility traces (location history) of both Bob and Alice. To make matching on this information

easier, Alice can provide each point in her GPS trace,

, as a geohash [

] at an arbitrary resolution,

Geohash converts GPS coordinates to strings at a varying length of string to provide an arbitration

resolution. For example, one can know that a GPS coordinate is inside the MIT Media Lab building

using the string ‘

drt2yr7x

’. If Alice desires, she can allow others to match on extremely detailed

location traces by using a high-resolution geohash in the form of a long string (

r= 8

characters). Or,

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：

BUILDINGAHEALTHIERFEED:PRIVATELOCATIONTRACEINTERSECTIONDRIVENFEEDRECOMMENDATIONSAPREPRINTTobinSouth1,2,*,NickLothian3,TakahiroYabe1,andAlex‘Sandy’Pentland1,21MITConnectionScience,MassachusettsInstituteofTechnology,Cambridge,USA2MITMediaLab,MassachusettsInstituteofTechnology,Cambridge,USA3Verida,Adel...

展开>> 收起<<

Building a healthier feed Private location trace intersection driven feed recommendations_2.pdf

共8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！

Building a healthier feed Private location trace intersection driven feed recommendations_2

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: