Impact of Search Engines on Page Popularity - Case Study Example

Describe the sources of bias that are found in search engine rankings and discuss the extent to which these may be commercially, politically or culturally important. Table of Contents Abstract I. Introduction II. SE ranking i. SE ranking a. User Query b. Factors Determining Query Scores ii. SE Page Ranking bias iii. Sources for SE ranking bias III. Impact of SE ranking bias i. Commercial impact ii. Cultural impact iii. Political impact iv. Social Impact IV. Conclusion V. Reference Abstract This paper focuses on Search Engine page ranking system. Various Page Ranking (PR) computing methods are evaluated to determine the source of bias in SE page ranking. Some of the reasons for SE bias are paid placements, provider’s choice of SE for free listing, SE search & retrieval algorithm and content provider methods. SE is motivated to introduce bias in web-page ranking to enhance its popularity. SE earns more revenue from paid listings & pay-per-click revenue. SE gains popularity by providing easy & free listing through incoming-link crawlers. Bias increases competition in providers with large content database. The content providers must not misuse the SE algorithm to gain popularity and high PR. While embedding links to popular sites provides free listing in SE database, unnecessary keyword stuffing is not ethical. The positive impact of page ranking system on web users & providers is the endurance to provide better service and quality to strive through the competition. This paper concludes that SE must be sensitive to user disutility with its page ranking policies. SE must define a bias threshold to maintain its credibility and improve quality of service to control user disutility. I. Introduction The bias created by the SE ranking system has commercial, cultural and political impact on the growth of businesses and social development. The objective of this paper is to research SE ranking, the sources of bias in SE ranking methods and the impact of these bias on various areas of users’ life socially, culturally, politically and commercially. The first section of the paper deals with SE Ranking which is covers the search engine ranking in general, SE biasness and sources of biasness. This section is followed by the impact of SE Ranking biasness on various areas like political, social, cultural and commercial followed by conclusion. II. SE Ranking Search engine ranking works on the keywords responses entered by the internet surfers. The search engine like Google, Yahoo and others list the pages matching with these key words and result in the list of webpage usually 10 pages on each search on their website. The search engine ranking is based on many factors like popularity of the website, payments to the search engine company by the listed company and others. This generates the ranking biasness as the new websites or the best information for the surfers’ response is not listed in the first pages of search results. This means the companies are using search engine as a tool to control the information need of the internet user and promoting their own ideas, products or service information rather than providing them a fair opportunity to get informed about the best options they have. i. SE ranking bias The size of the index-able web1 estimated at the end of January 2005 is at least 11.5 billion pages (GULLI and SIGNORINI, 2005:902). a. User Query SE Ranking is the rank of a web page in the list of web pages found in response to a user query. Based on this user query the search engine optimisers adopt specific strategy. Hence understanding of the types of users queries helps in understanding the Search engine ranking strategies. The user query can be of many types (KANG and KIM, 2004:64): Topic relevant Find Homepage Locate Service Topic relevant is the traditional and ad hoc way of information retrieval by user like user may enter “what is search engine optimisation?’. Find homepage is when user is interested in particular website and they try to search it directly through the keywords like “Oxford University website”. Locate service is where user directly enters the service or product it is looking for like “buy pizza from pizza hut”. Classifying query into any category or class is a challenging job as most of the times they are over lapping each other. b. Factors Determining Query Scores The query scores are influenced by number of factors. These factors are taken care of by the search engine optimiser while optimising any website in any search engine. The following factors are used to determine the query score in query type database2. Distribution of terms – multiple keywords/phrases used with AND, OR. E.g. hub and switch. Mutual dependence of words – phrase with more than one word. E.g. ‘weather forecast’ to know about the weather forecast. Usage rate in anchor text – exact match of query in anchor text. Query grammar – use of verbs, adverbs and adjectives. E.g. ‘low3 cost high-speed internet connection’, ‘selling4 old car’. ii. SE Page Ranking bias SE web-page ranking algorithms are based on various factors like probability, damping factors, false or spoofed page ranks and others. Page ranking based on sources of information: PR(A) = f(tf,df) where tf is ‘Raw frequency of a given term inside a document’ (KANG and KIM, 2004:65) and df is ‘Number of documents in which the index term appears’ (KANG and KIM, 2004:65). A web-site with high tf and df values will be ranked high. E.g. OKAPI scoring function (KANG & KIM, 2004:66) As a function of number of incoming links to page A from pages T1…Tn (page T points to page A) and outgoing links C(A) from page A. The number of visits to a web-page by random-surfers5 is more/less than visits by SE surfer. SE counts web-page visits through SERP only, SE results are thus biased. iii. Sources for SE ranking bias The primary reason of the Search Engine Ranking biasness is related to profitability. Marketers, products and service providers and various organisations want people to know about their organisation and product and services they offer. The primary motive in such cases is sales or revenue generation. SE PR bias is created by: Paid placements – SE is under obligation to provide better quality of service to content providers who pay a fee for web-site listing in SE database. Incoming links – Web-pages include links to popular web-sites for free listing in SE database. While this enhances popularity of these popular web-sites, it gives an opportunity to new/unknown web-sites to be found by SE crawlers. Figure 1: Popularity Bias created by incoming links (CHO and ROY, 2004:22) This figure shows how web-page popularity6 (X-axis) raises as number of incoming links (Y-axis) increase. Keyword stuffing – Frequency of query term ranks a web-page. Stuffing query term (keyword) in the content to raise PR is unethical. This is being practised for the better ranking in Search engines. Quality – Web-page quality can be measured in terms of web traffic and quality of incoming links. The caution is required when a low popularity page is promoted prematurely because of a new link from popular page (CHO, ROY & ADAMS, 2005:561-562) whereas at the same time some times the good new website does not get highlighted or promoted due to lack of association with any popular website or low popularity. III. Impact of SE Ranking Bias The websites are one of the primary sources of information for many internet users. People use this to fulfil their information needs related to entertainment, buying and selling products or services, finding jobs and work and many others. In certain cases it is very crucial for the internet user to get right and reliable information. For instance a student seeking admission in a good college may get information on number of colleges and it might get the information of average college on the first page of the search engine while he might be looking for the best colleges. i. Commercial Impact Paid placement fee reduces SE credibility and hence its market coverage and per-user profit. The content providers get “a higher relevance score, a featured listing, or perhaps even a guaranteed retrieval for certain search terms” (BHARGAVA and FENG, 2002:118). If γ is the placement fee and there are x free listings7, the additional revenue generated for SE is γ(1-x) (BHARGAVA and FENG, 2002:119). SE ranking guides and influences consumer; consumer may select a product/service from a provider with high-rank web-site thus further boosting the rank of this web-site. “Commercial content providers are interested in click-through and conversion rates—i.e., the likelihood that a search engine user will enter into a commercial transaction with the content provider” (BHARGAVA and FENG, 2002:118). SE search and retrieval algorithm is modified to suit these web-sites to generate revenue for SE. If β is the extent of bias for paid content providers and λ(β) is the positive effect of bias, then the valuation of this provider increases to V(θ)(1+ λ(β)) where V(θ) is the valuation of provider θ8 without paid listing (BHARGAVA and FENG, 2002:119). The commercial impact of SE paid placement strategy is driven by the relationship of β to λ(β). If ε9 is the elasticity of λ wrt β, when ε is low the SE revenue is limited. At this time SE is careful of its paid placement strategy, providers will not make paid placement and disregard to user utilities is a critical determining factor for M. When ε is high SE takes advantage of paid placements as provider benefits are far greater than the disutility10 users suffer (BHARGAVA and FENG, 2002:120). The guideline for SE is to remain sensitive to user disutility: The bias β for paid placements must be guarded by the user interests and not by ε. The commercial impact of search engine ranking biasness is on various parties like user or consumer, product or service provider who has good ranking but low quality product or services or a good service provider, search engine optimiser and the sales or revenue generated through this. Consumer may not get opportunity to reach for the best option of his needs in first instance while the best offer providing company will not be able to reach its prospects or target consumers with its offerings. It will impact the overall sales and trade of a particular business. ii. Cultural impact SE bias shifts value from product to the product information; the impact is on society financial and moral health. “Articles in the business press and data from commercial research firms suggest that paid placement strategies have a negative impact on a search engine’s perceived quality and credibility” (BHARGAVA and FENG, 2002:119). This may result in legal issues, loss of popularity and market share. SE must trade-off paid-placement with advertisements to act as a referee and fair arbiters of content relevance. If q11 is the perceived quality of SE by the user and c is the quality threshold value desired by the users before they use SE, then SE market coverage is defined as (BHARGAVA and FENG, 2002:119): M= q(1 − β (1 − x)) _ 1 c Therefore in an ideal SE with β=0 and x=1,12 the highest valuation user (θ=1) will use the SE when M > 0, i.e. q>c. This user will obtain the SE ranking and market share without any fee bias, solely based upon product performance, customer preference, web-site content evaluation and advertisements. Not many people spend time reading thousands of SERP, the paid placement is returned first and the probability that this service/product is preferred by the consumer is high. SE must contribute in social development by guiding the user to the best content. New web-site may not be added into SE database for months, paid placements are made by new businesses to gain commercial and political advantage. SE must guard against promoting paid placement bias for all content. Competition between service/product providers is hampered because SE bias is not a fair-deal. Paid placement rankings are the easiest way to influence the customer decision making. Consumer awareness of market is influenced by SE bias. iii. Political impact The search engine raking can have political impact. The websites expressing particular political interest may be promoted to create favourable conditions for a particular group or organisation based on SE Ranking. Particular political party may optimise its webpage for the promotional campaigns over others. Hence this has political impact as well. Popularity of popular pages rise and rich become richer with SE PR bias. iv. Social Impact The social impact of the search engine page ranking biasness is considerable as users many times do not know about the search engine biasness and rely on the results of the first effort. This means they do not receive best of the options they have and this also impacts their decision making processes. Biased search results on product information illustrate a general problem of considerable social importance. The Web is replacing the traditional repositories that individuals and organizations turn to for the information needed to solve problems and make decisions. (MOWSHOWITZ and KAWAGUCHI, 2002) IV. Conclusion The study in this paper concludes that both SE and web-content providers contribute to the SE bias. The major impact of SE bias is on consumer awareness; the new and high-quality service introduction is delayed. The biasness has impact on the social, cultural, commercial and political conditions. The commercial sides of SEPR are part of the revenue model of any search engine. Search engines may emphasise more on the advertisements and advertising links than the paid ranking and can consider other factors for the ranking which will provide users the best of the answers for their information needs irrespective of popularity of websites. The SE PR system must be a mix of paid placements, popularity measurements, user feedback, market research and source & content of information. V. Reference Bhargava K. Hemant. & Feng Juan. (2002) Paid Placement Strategies for Internet Search Engines, WWW Conference, Honolulu, 117-123. Cho Junghoo. & Roy Sourashis. (2004) Impact of Search Engines On Page Popularity, WWW Conference, New York, 20-29. Cho Junghoo., Sourashis Roy. & Adams E. Robert. (2005) Page Quality: In Search of an Unbiased Web Ranking, ACM, 551-562, Gulli A. & Signorini A. (2005) The Indexable Web is More than 11.5 Billion Pages, WWW Conference, Chiba, 902-903. Introna D. Lucas. & Nissenbaum Helen. Shaping the Web: Why the politics of search engines matters, The Information Scociety, (viewed October 2006). Kang In-Ho. & Kim GilChange. (2004) Query Type Classification for Web Document Retrieval, Information Processing and Management, 40(3), 459-478. Lawrence Steve. & Giles Lee C. (2000) Accessibility of Information on the Web, Intelligence, 11(1), 33-39. Mowshowitz Abbe. & Kawaguchi Akira. (2002) Bias on the Web, Communications Of The ACM, 45(9), 56-60. Read More