تشخیص تقلب در بررسی های آنلاین مصرف کننده
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی|
|17729||2011||13 صفحه PDF||49 صفحه WORD|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Decision Support Systems, Volume 50, Issue 3, February 2011, Pages 614–626
2. چگونه بررسی های مصرف کننده در طول زمان تحول می یابند؟
2.1. داده ها
جدول 1. خلاصه آمار
2.2. مرتبه (زمان نسبی) به چه معناست؟
2.3. محرکان بالقوه برای بررسی هایی که در طول زمان کاهش می یابند
شکل 1. میانگین امتیازدهی مصرف کننده در طول زمان (مرتبه).
2.4. چرا مضنونیم که ممکن است تقلب در بررسی وجود داشته باشد؟
شکل 2. ویژگی بازگشت-میانگین بررسی های آنلاین.
شکل 3. درصد امتیازدهی ها در طول زمان.
3. تجزیه و تحلیل های نظری: فرآیند خود-گزینی خالص یا ترکیبی از خود-گزینی با دستکاری؟
جدول 2. نیروی محرکه برای بررسی های آنلاین در طول زمان.
3.1. مدل: فرآیند خود-گزینی خالص که ابتدا، مصرف کنندگان با "امتیازدهی بالاتر" وارد می شوند
3.2. مدل: دستکاری خالص (با احتمال دستکاری کاهشی در طول زمان)
3.3. اگر هیچ "سوگیری امتیازدهی مثبتی هدایت شده توسط سوگیری خود-گزینی" و هیچ "دستکاری" ای وجود نداشته باشد
3.4. آزمون تجربی و بررسی ثبات
جدول 3. نتایج رگرسیون مورد انتظار با محرک های مختلف.
جدول 4. رابطه بین امتیازدهی ها و امتیازدهی متوسط تاخیری
4. رابطه بین کیفیت و دستکاری 9
شکل 6. رابطه بین کیفیت و امتیازدهی متوسط.
5. آیا مصرف کنندگان قادرند سوگیری را به طور کامل لحاظ کنند؟
شکل 8. انحراف بین متوسط امتیازدهی و کیفیت.
جدول 5. نتیجه تجزیه و تحلیل رگرسیون در مورد اینکه آیا مصرف کنندگان قادر به لحاظ کردن کامل سوگیری دستکاری هستند یا خیر (متغیر وابسته: ln(SalesRank)).
6. هنگامی که دستکاری وجود دارد، مشتریان چگونه تصمیمات خرید را اتخاذ می کنند؟
جدول 6. نتیجه تجزیه و تحلیل رگرسیون شاخص "کیفیت قیمت" (متغیر وابسته: ln(SalesRank)).
7. دستکاری در سراسر وب سایت ها
8. بحث ها، جمع بندی ها، و تحقیقات آینده
جدول A1. نتیجه رگرسیون در سطح مرتبه.
A.2. بررسی ثبات: مورد II
پانویس ها در متن
Increasingly, consumers depend on social information channels, such as user-posted online reviews, to make purchase decisions. These reviews are assumed to be unbiased reflections of other consumers' experiences with the products or services. While extensively assumed, the literature has not tested the existence or non-existence of review manipulation. By using data from Amazon and Barnes & Noble, our study investigates if vendors, publishers, and writers consistently manipulate online consumer reviews. We document the existence of online review manipulation and show that the manipulation strategy of firms seems to be a monotonically decreasing function of the product's true quality or the mean consumer rating of that product. Hence, manipulation decreases the informativeness of online reviews. Furthermore though consumers understand the existence of manipulation, they can only partially correct it based on their expectation of the overall level of manipulation. Hence, vendors are able to change the final outcomes by manipulating online reviewers. In addition, we demonstrate that at the early stages, after an item is released to the Amazon market, both price and reviews serve as quality indicators. Thus, at this stage, a higher price leads to an increase in sales instead of a decrease in sales. At the late stages, price assumes its normal role, meaning a higher price leads to a decrease in sales. Finally, on average, there is a higher level of manipulation on Barnes & Noble than on Amazon.
The rapid adoption of Web 2.0 has unleashed a wave of innovations that might change the way customers acquire information to make product purchases or stock investment decisions. The growth of Web 2.0 has enabled consumers to post reviews describing their experiences with products, product vendors, or service providers and make them available to other prospective consumers. In fact, the marketing literature suggests that consumers depend on online product reviews to make purchase decisions  and . Capital markets research has revealed that the information conveyed by stock message boards are used by investors , and a shock to the message board postings is negatively associated with future stock returns . Since consumers increasingly depend on information released through social online channels, such as consumer-generated content, to make product or services purchase decisions, the quality and truthfulness of information available to them is important. Do various entities, such as companies, vendors, publishers, or writers, actively engage in word-of-mouth manipulation, either directly or indirectly, with the goal of changing consumers' final decisions? Such practices are not new for information released through traditional information channels. For example, a rich earnings management literature has revealed that managers deliberately misrepresent financial reports in order to smooth their firm's income, meet a pre-specified target, and get better compensation. We define review fraud as occurring when online vendors, publishers, or authors write “consumer” reviews by posing as real customers. An email interview with Jonathan Carson, CEO of BuzzMetrics, reveals that promoting new CD releases through chat promotion is almost an industry standard . Such a practice exists even for highly reputable vendors, such as Amazon. In April 2004 James Marcus, a former senior editor for Amazon.com, wrote an alarming article in The Washington Post to discuss review fraud. Based on an analysis of reviews of just a few thousand reviewers, he found that a large number of authors on Amazon had got favorable reviews from their friends, relatives, colleagues or paid professionals. In some cases, these authors even wrote reviews for their own books. 1 Furthermore, such fraud has caused financial loss to society as well. 2 Recent research concludes that word-of-mouth (WOM) communication is a valuable marketing resource for consumers and marketers with critical implications for a product's success. This literature provides useful insights by linking online reviews with sales. It shows a positive correlation between the average review score and product sales ,  and . However, there is one implicit but essential assumption in this literature that researchers take for granted as being true, which is: Assumption 1. Online reviews are written by actual previous customers, not publishers or vendors, etc. Therefore, online reviews reflect either the actual product quality or the product's relative true quality. If the above assumption is true, then online reviews should reflect a products' true quality; or, all other information (e.g., price, product category, manufacturer, vendor, and shipping terms) being the same, a product with a higher mean consumer product rating should be assumed to have higher quality. This assumption is crucial in justifying the linkage between online reviews and sales. However, the existence of review fraud would invalidate such an assumption and cast doubts on the association between product quality and consumer reviews. If online reviews are indeed written by actual previous customers, then online reviews can help new customers reduce the uncertainties involved in inferring product quality, thus resulting in an increased conversation rate and higher sales. However, if online vendors, publishers, and authors are all able to write “consumer” reviews, then instead of being an uncertainty “reducer”, online reviews might become an uncertainty enhancer. In such a case, consumers' beliefs about product quality and vendor reputations derived from online reviews might be totally misleading. To date, there have been a few analytical studies investigating review fraud  and . Drawing on the observation that the music industry is known to hire professional marketers to write favorable consumer opinions to promote the sales of new albums, Mayzlin  built an analytical game theory model in which two competing firms send anonymous messages recommending their own products. Dellarocas  analytically shows that if every firm's manipulation strategy monotonically increases with regard to that firm's true quality, then manipulation of online reviews increases the informativeness of online reviews. Under such a circumstance, manipulation increases the separation of the distributions of ratings and will help consumers make better purchase decisions. Even if there is manipulation, consumers are smart and can adjust their interpretation of online opinions accordingly . Combining the implicit assumption stated above (Assumption 1) with these analytical works, we have the following revised assumption based on previous literature: Assumption 2. Online reviews are written by actual previous customers and not publishers or vendors. Even if there is manipulation, consumers are smart and can adjust their interpretation of online opinions accordingly . Further, as long as the manipulation is monotonically increasing with regard to a product's true quality (i.e., if it is more likely for higher quality vendors to engage in review manipulation), then online reviews with the existence of review fraud are even more informative than when there is no review fraud. If consumers are indeed smart and if the manipulation is monotonically increasing with respect to (w.r.t) to product quality, then we need not worry about empirically testing manipulation of online reviews because under such a circumstance, online reviews are more informative. However, are these assumptions true? In this paper, we analytically and empirically study temporal behaviors of online reviews and address the following research questions: • Does review fraud actually exist? Is review manipulation a prevalent phenomenon or does it just happen occasionally? • What types of vendors are more likely to manipulate online reviews: those selling high-quality products or those selling low-quality products? Vendors that receive higher average ratings for their products, or those with lower average ratings? • Are consumers smart enough to filter out the manipulation as Dellarocas  suggests? Are they able to correct for this bias in their purchase decisions? What quality indexes do they use to make purchase decisions in view of the existence of review fraud? • Is online review fraud a common phenomenon across different websites? This paper proceeds as follows. Section 2 studies the mean-reverse phenomenon of consumer reviews to motivate our study. By studying the temporal patterns of online reviews, we show that there might be two potential drivers which are consumer taste difference and/or review manipulation that force rating decreases over time. As a nature follow-up question of 2 and 3 answers whether a pure consumer taste difference without manipulation can be the sole underlying driving force. We conclude that we cannot rule out manipulation as one of the potential drivers. The temporal patterns of online reviews can be either driven by pure manipulation or by a joint force of consumer taste difference and manipulation. Section 4 seeks to answer the question of whether low-quality or high-quality vendors are more likely to manipulate consumer reviews. Section 5 analyzes whether consumers correct for manipulation bias when making purchase decisions. Section 6 answers how customers make purchase decisions when manipulation exists. Section 7 checks the robustness of our findings by comparing the online review manipulation between Amazon and Barnes & Nobel. Section 8 contains discussion of the findings, their implications, and some concluding remarks.
نتیجه گیری انگلیسی
In this study, we use data from Amazon and Barnes & Noble to document that publishers, authors, and vendors consistently manipulate online consumer reviews. If a firm decides to adopt manipulation, its manipulation strategy is monotonically decreasing with respect to that product's true quality. Under such a case, manipulation actually decreases the informativeness of online reviews. However, we prove that not all firms will manipulate online reviews. Because of this non-systematic involvement, it is not easy for consumers to fully correct for manipulation bias. Consumers can adjust for that bias based only on their expectations about overall manipulation rates. To some degree, vendors are able to manipulate the outcomes of the results and consumers therefore respond to the wrong information. We document the existence of the “price quality proxy” in the sense that at the early stage after an item is released to the Amazon market, consumers use price as a quality indicator instead of using the average rating. Thus, a higher price leads to an increase rather than a decrease in sales. Finally, we show that generally there is a higher level of manipulation on Barnes & Noble than on Amazon. We document that the lower the quality and average rating of the products a vendor is selling, the higher the likelihood that that vendor is going to conduct online manipulation. This makes online reviews much less informative than when either there is no manipulation or when vendors selling higher quality products are more likely to manipulate online consumer opinions. This might result in consumers' totally discarding online reviews, defying the purpose of vendors' building online review systems and providing customers with an online review option. Over the long run, online markets such as Amazon.com or Barnes & Noble.com cannot maintain the quality of their online consumer opinion information when such manipulation is taking place. If the market continues to evolve in this way, customers will no longer read these online reviews. We urge the key players in these online markets to find a way to increase the cost of manipulation in order to mitigate the manipulation effect. We call for collective thinking within this community, including the technical vendors and business entities, to build a better online system to fight against this practice. The ideal situation would be that online reviews represent the truth, the whole truth, and nothing but the truth about their products. However, unless we can resolve the manipulation issue, online consumers can only get the “partial truth.”