Quantifying Political Bias in News Articles

2025-05-02 0 0 1.27MB 9 页 10玖币

侵权投诉

antifying Political Bias in News Articles

GIZEM GEZICI, Huawei Turkey R&D Center, Turkey

Search bias analysis is getting more attention in recent years since search results could aect In this work, we

aim to establish an automated model for evaluating ideological bias in online news articles. The dataset is

composed of news articles in search results as well as the newspaper articles. The current automated model

results show that model capability is not sucient to be exploited for annotating the documents automatically,

thereby computing bias in search results.

Additional Key Words and Phrases: Bias evaluation, Fair ranking, Political bias, Web search

ACM Reference Format:

Gizem Gezici. 2022. Quantifying Political Bias in News Articles. 1, 1 (October 2022), 9 pages. https://doi.org/

10.1145/nnnnnnn.nnnnnnn

1 INTRODUCTION

Search engines are ubiquitous. As reported by SmartSights [

], in 2017 46.8% of the world population

accessed the internet and by 2021, the number is expected to reach 53.7%. Currently, on average

3.5 billion Google searches are done per day [

]. These statistics advocate that search systems

are “gatekeepers to the Web” for many people [

]. As information seekers search the Web more,

they are also more inuenced by Search Engine Result Pages (SERPs) and their inuence -negative

included- potentially become visible in a wide range of areas. However, as with all software-based

systems, search platforms do not lack of human inuence, thereby they may suer from embedded

bias, i.e., corpus or algorithmic bias. Experimental studies suggest that particular information types

or sources might be retrieved more or less, or might not be well represented [

]. For instance,

during the elections, it is known that people issue repeated queries on the Web about political

candidates and events such as “democratic debate”, “Donald Trump” and “climate change” [

Epstein and Robertson

[7]

claim that SERPs returned in response to these queries may inuence the

voting decisions of the users and report that manipulated search rankings can change the voting

preferences of undecided individuals at least by 20%. As individuals rely to a greater extent on the

SERPs for their decision making, there is a thriving demand for these systems explainable.

Empirical evidence has shown that individuals trust more sources ranked at higher positions

in SERPs, but the ranking criteria may rather depend more on user satisfaction than the factual

information, which jeopardizes the phenomenon of providing reliable information in exchange

for satisfying users [

]. In spite of the given research ndings, the majority of online users tend

to believe that search engines provide neutral results, i.e., serving only as facilitators in accessing

information on the Web due to their automated operations [

]. However, this romanticised view

of search platforms does not reect reality and there seems to be a growing skepticism related

to objectivity and credibility of these platforms. To illustrate that, a recent dispute between the

U.S. President Donald Trump and Google can be given, where Mr. Trump accused Google of

Author’s address: Gizem Gezici, Huawei Turkey R&D Center, Istanbul, Turkey, gizem.gezici@huawei.com.

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee

provided that copies are not made or distributed for prot or commercial advantage and that copies bear this notice and

the full citation on the rst page. Copyrights for components of this work owned by others than ACM must be honored.

Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires

prior specic permission and/or a fee. Request permissions from permissions@acm.org.

XXXX-XXXX/2022/10-ART $15.00

https://doi.org/10.1145/nnnnnnn.nnnnnnn

, Vol. 1, No. 1, Article . Publication date: October 2022.

arXiv:2210.03404v1 [cs.IR] 7 Oct 2022

2 Gezici, et al.

presenting only negative news about him when his name is searched. Google refuted this claim by

saying that: “When users type queries into the Google Search bar, our goal is to make sure they

receive the most relevant answers in a matter of seconds” and “Search is not used to set a political

agenda and we don’t bias our results toward any political ideology”

. In this work, we hope to

shed some light on that debate, by not specically examining queries regarding Donald Trump but

by fullling an in depth analysis of retrieved search answers to a broad set of queries related to

controversial topics based on concrete evaluation measures.

Bias implies undue emphasis. For a retrieval system, it can be dened as the balance and repre-

sentativeness of Web documents retrieved from a database for a set of queries [

]. When a user

issues a query to a search engine, documents from dierent sources are collected, ranked, and

presented to the user. Assume that a user searches for 2016 presidential election and the top-n ranked

results are displayed. In such a search scenario, the retrieved results may exaggerate or downplay

particular perspectives and thereby provide an unbalanced picture of the given query as claimed by

Mr. Trump, though without any scientic support. Hence, the potential undue inclusion or exclusion

of specic perspectives in the retrieved results lead to bias [

]. Note that the existence of bias is

dierent from relevance. In the presented scenario, even though the retrieved list of documents

may all be judged as relevant with respect to the given query, if the selection of the documents is

skewed or slanted, i.e., emphasizing one perspective over another, then the corresponding search

engine is biased due to an imbalanced representation of the perspectives towards the query’s topic.

Bias is especially important if the query topic is controversial having opposing perspectives as

described in the given scenario. The bias in SERPs can be used by search engines to inform their

users about the bias by making themselves more accountable which is one of the crucial attributes

that a retrieval system should possess [4].

In this work, we focus on the SERPs coming from the news sources and investigate two major

search engines (Bing and Google) in terms of political bias. Our analysis has mainly three sides

where we evaluate the political bias of the search engines separately, then make a comparison

among them as well as track the source of this bias in the SERPs. The bias may come from the

data, which may contain biases (input bias) or the search algorithm, which contains sophisticated

features (algorithmic bias). In the scope of this work, we concentrate on the input bias that is

intrinsically embedded in the data itself.

In short, we aim to answer the following research questions:

RQ1:

On a conservative-to-liberal scale, do search engines return politically biased SERPs in

response to queries related to controversial topics?

RQ2: Are search engines signicantly dierent from each other towards controversial topics?

RQ3: Does the source of bias come from the input data?

We address these research questions for controversial topics representing a broad range of issues

in SERPs of Google and Bing through content analysis, i.e. analysing the textual content of the

retrieved documents. We focus on content bias and describe our SERP politically content bias

quantication framework in which we propose three dierent measures of bias based on common

Information Retrieval (IR) utility-based evaluation measures: Precision at cut-o (P@n), Rank Biased

Precision (RBP), and Discounted Cumulative Gain at cut-o (DCG@n). While the rst measure

quanties bias considering only a weak ranking criterion, i.e. the rst

𝑛

returned documents as in

SERPs, the other two measures incorporate stronger ranking bias.

In order to answer RQ1, we measure the degree of deviation of the ranked SERPs from an

ideal distribution where dierent political perspectives are equally likely to appear [

]. To detect

https://www.reuters.com/article/us-usa-trump-tech-alphabet/google-responds-to-trump-says-no-political-motive-in-

search-results-idUSKCN1LD1QP

, Vol. 1, No. 1, Article . Publication date: October 2022.

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：

QuantifyingPoliticalBiasinNewsArticlesGIZEMGEZICI,HuaweiTurkeyR&DCenter,TurkeySearchbiasanalysisisgettingmoreattentioninrecentyearssincesearchresultscouldaffectInthiswork,weaimtoestablishanautomatedmodelforevaluatingideologicalbiasinonlinenewsarticles.Thedatasetiscomposedofnewsarticlesinsearchresult...

收起<<

Quantifying Political Bias in News Articles.pdf

共9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！

Quantifying Political Bias in News Articles

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: