Has anyone summarized or meta analyzed the gains from micro targeting and intervention personalization in any domain?
Working on my spring syllabus for the design/governance of field experiments. cc @deaneckles @sinanaral
Working on my spring syllabus for the design/governance of field experiments. cc @deaneckles @sinanaral
@deaneckles @sinanaral Here's a fascinating thread by @garjoh_canuck showing that micro-targeting increases ad prices, which means that advertisers at least think it's valuable?
@deaneckles @sinanaral @garjoh_canuck In a quasi-experiment on the effectiveness of targeted advertising, @avicgoldfarb & @ce_tucker found that government cookie restrictions dramatically reduced the effect of online banner ads on purchase intent (from 0.03 to 0.002) pubsonline.informs.org/doi/abs/10.128… 
@deaneckles @sinanaral @garjoh_canuck @avicgoldfarb @ce_tucker Even when people increase an *intent* to buy or vote, an intervention may still have little effect on behavior. That's what Kyle Endres & @professorcostas found in a recent experiment targeted to voters whose policy preferences differed from their party journals.uchicago.edu/doi/full/10.10…
@deaneckles @sinanaral @garjoh_canuck @avicgoldfarb @ce_tucker @professorcostas Overall, what is the effect of campaigning and advertising on candidate choices in US elections? @j_kalla and @dbroockman, after reviewing 40 field experiments, argue that the average effect is *zero* papers.ssrn.com/sol3/papers.cf…
@deaneckles @sinanaral @garjoh_canuck @avicgoldfarb @ce_tucker @professorcostas @j_kalla @dbroockman As @Adam_Schaeffer & @sinan pointed out, Kalla & Broockman find that campaign interventions *can* persuade people when campaigners do cycles of tests to find & target specific "persuadable voters," an approach which requires more than just the typical data used in targeted ads
@deaneckles @sinanaral @garjoh_canuck @avicgoldfarb @ce_tucker @professorcostas @j_kalla @dbroockman @Adam_Schaeffer @sinan So far in this thread, I've been sharing evidence on the effects (or non effects) of targeted ads. They're often paired with tailored or personalized messaging. Do such messages work? @eitanhersh & @b_schaffner found that voters prefer wider appeals jstor.org/stable/10.1017…
@deaneckles @sinanaral @garjoh_canuck @avicgoldfarb @ce_tucker @professorcostas @j_kalla @dbroockman @Adam_Schaeffer @sinan @eitanhersh @b_schaffner In this thread, I asked for evidence on effects of targeted & personalized ads. Political scientists shared links on influencing voting behavior, an area of uncertainty where the stakes are high. Here's @sinanaral & @deaneckles argument for more research sciencemagazinedigital.org/sciencemagazin…
@deaneckles @sinanaral @garjoh_canuck @avicgoldfarb @ce_tucker @professorcostas @j_kalla @dbroockman @Adam_Schaeffer @sinan @eitanhersh @b_schaffner What about personalization? Have there been meta-analyses on the effects of personalizing interventions in any domain? (marketing, health, political science, anything)
@deaneckles @sinanaral @garjoh_canuck @avicgoldfarb @ce_tucker @professorcostas @j_kalla @dbroockman @Adam_Schaeffer @sinan @eitanhersh @b_schaffner @swattal, Telang, Mukhopadhyay, & @peterboatwright
ran non-causal correlations on millions of email ads & found "consumers respond negatively when firms are explicit in their use of personally identifiable information" including personalized greetings pubsonline.informs.org/doi/10.1287/is…
ran non-causal correlations on millions of email ads & found "consumers respond negatively when firms are explicit in their use of personally identifiable information" including personalized greetings pubsonline.informs.org/doi/10.1287/is…
@deaneckles @sinanaral @garjoh_canuck @avicgoldfarb @ce_tucker @professorcostas @j_kalla @dbroockman @Adam_Schaeffer @sinan @eitanhersh @b_schaffner @swattal @peterboatwright Contextual bandits are commonly used to personalize all sorts of things, including the artwork shown to individual viewers on Netflix. Are there any meta-analyses of the effectiveness of this approach? cc @justinbasilico
medium.com/netflix-techbl…
medium.com/netflix-techbl…
@deaneckles @sinanaral @garjoh_canuck @avicgoldfarb @ce_tucker @professorcostas @j_kalla @dbroockman @Adam_Schaeffer @sinan @eitanhersh @b_schaffner @swattal @peterboatwright @JustinBasilico 😱🤣You know those ads with products you previously viewed? "dynamic retargeted ads are on average less effective than their generic equivalent" (unless advertisers detect you're about to change your mind)
Field experiment by @anja_lambrecht & @ce_tucker journals.sagepub.com/doi/10.1509/jm…
Field experiment by @anja_lambrecht & @ce_tucker journals.sagepub.com/doi/10.1509/jm…
@deaneckles @sinanaral @garjoh_canuck @avicgoldfarb @ce_tucker @professorcostas @j_kalla @dbroockman @Adam_Schaeffer @sinan @eitanhersh @b_schaffner @swattal @peterboatwright @JustinBasilico @Anja_Lambrecht Is there really little evidence on the effects of *personalized* ads? Computer scientists have published papers on algorithms but I haven't found meta-analyses on personalization. Is that because they're so hard to compare to each other?
@deaneckles @sinanaral @garjoh_canuck @avicgoldfarb @ce_tucker @professorcostas @j_kalla @dbroockman @Adam_Schaeffer @sinan @eitanhersh @b_schaffner @swattal @peterboatwright @JustinBasilico @Anja_Lambrecht Or am I just using the wrong terms? Here, for example is a meta-analysis of "computer-tailored" messages in health behavior change, where they appear to combine group-level custom messages with more personalized interventions ncbi.nlm.nih.gov/pmc/articles/P…
