Social Impact IDentification (SIID)

Overview

Ads with social bias have a negative effect on everyone. Companies lose millions and millions of dollars. Employees and leads are fired. Audiences lose their trust and, sometimes, gain trauma. In fact, according to Ad Week, more than one-third of Americans will refuse to purchase from a brand because of bad advertising - even if the ad is changed. Some recent examples seen in the past 3 years were Gucci’s blackface sweater, Heineken’s “Sometimes, Lighter is Better” commercial, and Pepsi’s protest commercial with Kendall Jenner. 

Eri O’Diah came to my team and I with a meaningful, impactful challenge to research and design a digital tool called the Social Impact IDentification, or SIID for short, that leverages artificial intelligence to detect and prevent social bias in marketing campaigns. As a founder of a marketing agency, she has seen first hand how biased marketing campaigns have resulted in losses and damages, lost consumer trust in a brand, and negative social impact.

The User Groups

Primary Users

1. Marketing and advertising specialists who wish to strategize and execute effective campaigns that reach their target audience(s)

2. Visual designers looking to leverage visuals to communicate the marketing/design strategy to their target audience(s)

Secondary Users

1. Marketing directors or organizational leaders who wish to create maximum exposure to their brand while limiting risk to their brand value and meeting business objectives.

Research

Discovery Research

To tackle this complex problem space, my team and I researched the following areas of focus to build a foundation of understanding:

• Marketers and their work with Ad Campaigns (my focus area)

• AI-assisted Web Applications (my focus area)

• Artificial Intelligence

• Social Bias

We used a combination of interviews with subject matter experts, secondary research, online surveys, and competitive audit with other AI-assisted web applications that work with documents such as Grammarly, Textio, Hemingway Editor, and Boomerang Insights.

User Testing

After each round of prototyping (there were 2 rounds of prototyping before our finalized prototype), we evaluated our early to mid-level fidelity prototypes with in-person or remote user test sessions with the following protocols:

1. Research protocol with speed dating elements to test early web app feature concepts. (See my protocol here.)

2. Storyboard research protocol to test bias expert consultation and training concepts.

3. Research protocol with A/B testing elements to test mid-fidelity web app prototypes.

Synthesis

Personas

Based on our discovery research, I synthesized the research findings regarding SIID’s key user groups by creating these personas to gain empathy for the users and better understand their needs, preferences, and frustrations.

User Journey Map

I also created a user journey map of a marketer’s journey during an ad campaign to gain empathy for the user group’s experience and better understand the pain points and opportunities during this process. 

Affinity Diagramming

After my team collectively finished collecting data from 10 individuals, we used affinity diagramming (KJ Method) to determine key themes and insights from our users. Using the personas, user journey map and insights from our affinity diagramming, we designed our high fidelity prototypes.

The Prototype

Main Features

Bearing in mind who the users are and their process with ad campaigns (see User Journey Map above), our team developed a powerful web application platform to combat negative social impact. The application includes these 3 features:

Rationale for Features Based on Key Research Insights:

1. Marketing Process: We learned that marketers work in a quick-paced environment, often juggling multiple ad campaigns with tight deadlines and budget constraints. Our user feedback during testing revealed that marketers highly value digital tools that can help them manage several tasks at the same time.

2. Social Impact Detector: Users responded more positively to flagged alerts for social bias in comparison to an overall score (scale, percentages, etc.) of bias, which users did not fully trust. Users also preferred brief annotations with explanations of why a term is biased because it makes alerts more educational and actionable.

3. Bias Training and Education: Based on research and user feedback, our team determined that, no matter how sophisticated the AI, real change could only happen with people. Given its limitations, AI cannot be a standalone solution for bias. Users expressed interest in having a human aspect, in which they can consult with a bias expert to learn more about their biases.

The Interactive Prototype

My team and I built a high-fidelity, interactive prototype of the SIID web application using Figma. Below is a screenshot of the web application’s Social Impact Detector feature that I prototyped, as well as a short video that demonstrates the SIID prototype walkthrough (on the right). To view and interact with the prototype directly in Figma, click on the link here.

Next Steps

A future concept we considered is the SIID Plug-In feature that I prototyped in Figma (as featured below). This plug-In would be similar to the web application in that it would be able to identify potential bias. It would be compatible with a variety of browsers and would scan the content on any open page to identify or highlight terms that are at risk for bias.

A user would not be able to upload a document like the web application, but this tool would seamlessly work alongside many of the programs and tools that marketers often use, (like Google docs, email, Hubspot and others). As AI technologies around bias and the awareness of SIID continue to grow, the Plug-In feature would be a great opportunity to pursue.