Leave a comment


Mar 7, 2011 | Discerning dated Google results

In one of the classes I am taking this quarter, we are investigating ways in which social norms propagate through the web, particularly with social networks. Social norms are the sort of “standards” that we subconsciously absorb by observing cues like how other people behave. If on a video you see a thread of comments that looks unfavorably upon a video, your perspective of the video is likely to be slightly impacted.

What I wanted to investigate as a tangent to that was ways in which the search engine paradigm really is affected by those unnamed individuals around us online. A website about shoes, for example, can only make its way to the top of a search engine if lots of people consider that site to be useful regarding shoes. In short, your search efforts are made easier because lots of people also found it to be the best solution.

Some search results return a timestamp that denotes when the particular page had its content written. This is particularly true for blogs or news sites. I wondered if in the first page of results people would go for the most recent page, even though it wasn’t ranked as highly. This isn’t a scientifically rigorous test by any means – I used 72 individuals on FiveSecondTest to click on the first link that they would if they were searching for the best photo-sharing website. Each bucket of 24 individuals was given a different screenshot. One was the original Google search result, the second put the oldest link (2005) on the top and a couple newer ones below, and the third one put a less old article (2008) on top, (2005) below and newer ones below. Each of the screenshots have been paired with their clickmap counterparts.

Original query


Oldest one first


New but not newest first


There are a few problems with this study that can be improved upon in next iterations. 1) Not all the links on the pages are the same across all 3 conditions and 2) the first link does actually have a date on it (in the original query). Nonetheless, the results are fairly interesting: people take enough time to scan through the dates and click on the newest ones. This is particularly insightful if you think that any link on the first page is probably as popular and relevant as it could be to the query – any link could technically be as useful as another. If this same paradigm can be proposed for not only recency, but from other social cues (like 10 out of your 50 friends clicked on this link) – how different will the search engine landscape be?

This entry was posted on Monday, March 7th, 2011 at 2:04 am, EST under the category of Coding. You can leave a response, or trackback from your own site.