An experimental comparison of click position-bias models

By Nick Craswell, Onno Zoeter, Michael J. Taylor, Bill Ramsey. 2008

1,075 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Information Systems

Citations in the corpus

Offline A/B Testing for Recommender Systems (2018, background): “based on the interests of the user. This task is formalised as a ranking task, and not only a top-K retrieval, because the different product slots are not equivalent and exhibit different performance [5]. The system outputs a ranked list of products and then maps better products to better positions.”
A Tutorial on Thompson Sampling (2017, methodology): “We consider formulating this problem as a cascading bandit, in which user selections are governed by a cascade model, as is commonly used in the fields of information retrieval and online advertising [28].”
How Visibility and Divided Attention Constrain Social Contagion (2012, background): “Visibility not only impacts retweet selection, but it helps explain oth er observations, such as click position bias [37], [49].”
Dynamics of Content Quality in Collaborative Knowledge Production (2017, background): “2012), and the role of cognitive heuristics in information search and retrieval (Craswell et al .”
Explore-exploit in top-N recommender systems via Gaussian processes (2014, background): “The position dependence p can often be estimated effectively [6].”
Bias in Online Freelance Marketplaces: Evidence from TaskRabbit and Fiverr (2017, background): “Prior work has shown that status differentiation and placement in rankings do impact human interactions with online systems [49, 18], which suggests that similar effects will occur on online freelance marketplaces, but we lack the data to empirically confirm this.”
Equity of Attention: Amortizing Individual Fairness in Rankings (2018, background): “The existence of position bias in rankings of search results has been revealed by a number of eye-tracking and other empirical studies [10, 11, 19].”
Click Models for Web Search (2015, methodology): “e cascade model [Craswell et al., 2008] assumes that a user scans documents on a SERP from top to bottom until they find a relevant document. Under this assumption, the top ranked document u1 is always examined, while documents ur at ranks r 2 are examined if, and only if, the previous document ur ”
Beyond Ranking: Optimizing Whole-Page Presentation (2016, background): “Eye-tracking experiments [20, 19] and click log analyses [24, 15] observe that users follow sequential order in browsing blue-link-only SERPs.”
Relative confidence sampling for efficient on-line ranker evaluation (2014, methodology): “To model the user’s click behavior on the resulting interleaved lists, we employ a probabilistic user model [9, 14] that uses as input the manual labels (classifying documents as relevant or not for given queries) provided with both datasets.”
Mining Query Logs: Turning Search Usage Data into Knowledge (2010, background): “[66] have presented a more thorough elaboration bias due to presentation order.”
Expected reciprocal rank for graded relevance (2009, background): “Position-based models [12, 20] are a popular class of meth-ods for dealing with the presentation bias problem inherent in ranked retrieval systems.”
Transfer Learning for Behavior Ranking (2017, methodology): “Notice that the seemly very similar click models in information retrieval such as the cascade model [Craswell et al. 2008], position model [Chapelle and Zhang 2009], and general click model [Allen Zhu et al. 2010] actually differ: (i) they do not answer the above ﬁrst question, because they require…”
Learning to Rank with Selection Bias in Personal Search (2016, methodology): “[22] also proposed five simple strategies to eliminate these biases, including the Click > Skip Above strategy, which gave rise to the well-known Cascade model [11].”