An Investigation into the perception of social media endorsement posts in accordance with the attributes of the endorser, utilising a chain of cultural transmission reminiscent of the Instagram algorithm.

Abstract

The following study is an investigation into social media product endorsements and how the individual promoting a product impacts how likely the post is to receive engagement from consumers. An experiment was created featuring simulations of influencer marketing posts one might see on Instagram, including a product image and profile picture. Some of the profile pictures contained variation in attractiveness, some posts contained a follower number, and some contained an occupation, or a combination of these attributes. The experiment was run via Amazon Mechanical Turk and a chain of cultural transmission was created as the first group of participants responses were collected and used to determine how frequently certain attributes would appear in the next group of social media posts shown in the experiment. The attributes that were liked most frequently in the previous set of posts were then expressed at a higher frequency for the next group of participants. This transmission chain element is unique to this study and aims to investigate whether there is a preference for attributes that are more common over attributes that are rare, while replicating the effect of social media algorithms that promote posts we have not seen before but were popular amongst others. The study produced complex results suggesting that there is no consistent effect of the frequency of representation of certain attributes on the chances of a post being liked. There was a significant preference for higher attractiveness profile pictures across all conditions and a preference for higher popularity when attractiveness was controlled. Additionally, there was a greater like frequency for relevant experts when presented alongside popularity but no significant preference for either relevant or irrelevant experts when presented alongside variation in attractiveness. Finally, there was no consistent effect of participant self-esteem on attribute like frequencies.