PowerMeter

Every time you enter a search query on Google search or Bing search the extension gets activated. Before displaying the search results it decides whether or not to perform a small modification to how results are displayed. This decision is completely randomized. A modification would correspond to hiding an ad and/or swapping the position of two randomly selected search results. The rest of the page is untouched and search results are never removed from the results page, at most they are moved up or down by a few positions. After displaying the results, the extension observes which index the user ended up clicking on and records the experimental configuration.

After installing the extension you can use your web browser as usual. Our goal is to measure the impact of search page display on user clicks within the context in which web search is used on an everyday basis. There is nothing special you need to do.

Powermeter only collects information that is strictly necessary for the experiment and in a native completely anonymous fashion. The collected data is related to: (i) the index of the clicked item in the list of search results, (ii) the relative positioning of the clicked item, and (iii) the index of the search page and the number of ads and specialised search reults present on the page the click occurs. Powermeter does not have access to users' Google account information. Powermeter does not store IP addresses, individual search queries, nor any information concerning the websites visited by users as result of their search. Your privacy is important to us. For more details, check our Privacy Policy.

Powermeter sends recorded information via https to a secure webserver hosted locally by the Max Planck Institute for Intelligent System at its premises in Tübingen, Germany. Access to the server is restricted and controlled.

Powermeter will automatically stop collecting data after the experiment is terminated. If participants decide to opt out earlier, then data collection will be stopped at the moment when they deactivate the extension. The planned duration of the experiment is until February 2024. However, researchers reserve the rights to adjust the duration depending on users active engagement. The status of the experiment will be documented on this website.

You can deactivate Powermeter at any time by removing the extension form your Chrome browser. Therefore, go to 'Settings' in the Chrome browser menu, select 'Extensions' and click 'Remove' next to the Powermeter logo. The extension will be deinstalled, removed form your browser and data will no longer be collected.

Collected data will be used for research purposes only, it will not be shared, distributed, or repurposed. Collected data will be used to compute aggregate statistics on click behavior. These statistics will then be used to draw inferences about the causal effect of positioning and display on users' click behavior. The goal is to publish the findings as part of a scientific paper.

The collected data will be aggregated and evaluated by authorized researchers participating in the research project. The findings will be published as part of a scientific paper. As soon as relevant statistics are elaborated, collected data will be deleted. Data will not be stored or otherwise repurposed.

The project is lead by Celestine Mendler-Dünner (Group Leader at the Max Planck Institute for Intelligent Systems), in collaboration with Moritz Hardt (Director at the Max Planck Institute for Intelligent Systems) and Gabriele Carovano (former Postdoc in AI law at University of Tübingen, now with the Italian Competition Authority).

The experiment has been stopped, as of extension version Powermeter v1.3. A summary of the results from the experiment periode September 2023 until January 2024 can be found on Arxiv