AMT Lab @ CMU

View Original

Navigating the Ethics of Using AI for Donor Solicitation

Written by Duncan Bindbeutal

AI has been slowing encroaching on every facet of modern life. It runs our phones, it navigates our vehicles, and has even began finishing our sentences. As of now, its integration into our everyday life has been mundane. Since AI operates most effectively behind the scenes and is difficult to explain to the average consumer, most products that utilize it do not necessarily draw much attention to it. Negative connotations among the public may prevent business from directly or overtly acknowledging the use of AI in their products or administration, but it certainly does not stop them from utilizing it. AI is bolstering many forms of business management technology, and the nonprofit sector is starting to reap its benefits. With so much of nonprofits reputations relying on trust and good faith connections between stakeholders, donors, and staff, what happens when AI is introduced into the dynamic?

AI Explained

Simply put, AI is the simulation of human thought processes by computer systems. It can be used for many things, including natural language processing, speech recognition, or expert systems (a simulation of judgement.) When computer systems utilize AI, they circumvent needing to have a rigid set of responses to specific inputs and may instead “learn” how best to process or interpret the data provided to them.

While there is no shortage of AI being used for novel purposes, its real value can be found in its aptitude for performing repetitive, detail-oriented tasks. However, it can be cost prohibitive and often requires a great deal of technical expertise to use, but new advances are made every day and consumers are getting access to more and more powerful programs.

Using AI in the nonprofit sector

There is much to be gained from bolstering a workplace of any size with this type of technology. If implemented effectively, it can allow people to reallocate time, money, and resources on mission-based issues and leave rote administrative tasks to thinking machines. AI has the potential to act as a force multiplier for any number of nonprofits by enabling them to accomplish measures more than they would be able to without it. However, when we examine how AI could most effectively be used in the Nonprofit sector, we run into some morally questionable areas quickly.

Fundraising is a major part of any Nonprofit organization, and email-based solicitations made 28% of all online nonprofit revenue in 2017, earning an estimated $43 dollars per thousand emails sent. If AI was effectively used by a Nonprofit to quickly generate text-based solicitation material designed for hyper specific donor segments, its stands to reason there is a lot more money to be earned. An established nonprofit would have a lot of this data on hand in the form of previously implemented donor solicitations. By providing all their previously written material to an AI program and instructing it to generate new material based on set parameters, the solicitation output of a Nonprofit could increase astronomically.

Generation Text

Text generative AI is a booming industry. There is a wide array of both free and subscription-based consumer and professional grade text generating AI, available to anyone with access to a web browser. OpenAI, a leading AI research and development company, has released three versions of their publicly available text generative AIs entitled GTP. They initially refrained from making their second version publicly available for concern that it would be used to contribute to the fake news cycle during the 2016 presidential election. Eventually the full version was released publicly, and the third iteration, GPT-3, was even enlisted to write a real news article for The Guardian. The article required an editor, which was generated based on 8 similar articles. Regardless, this demonstrates just how close we are to being able to automate this foundational segment of solicitation work.

To illustrate the relative ease in which workable donor copy could be generated from existing material, I fed some donor solicitation materials from the Mattress Factory, which was composed of 78 words, into the demo version of InferKit, a browser-based text generation tool. The results of the AI-generated content is written below in italics.

Donate today and make a difference tomorrow! When you donate to the Mattress Factory, you give artists the time, space and resources to create remarkable works of art that help us see our world in new ways. You support our education and outreach programs allowing young people and community members to imagine and explore a life of creative practice. You sustain a first-class collection of contemporary installation art and share it with nearly 100,000 museum goers every year. That's priceless! Your gift also supports the Mattress Factory's operating budget, our educational programming, and the creation of public events. You are helping to turn the arts and culture world on its head by creating more pathways for artists and inspiring a new generation of culture changers.

What is your contribution supporting?

The Mattress Factory provides artists with the time, space, and resources they need to produce imaginative work that encourages social change. Our public programs help connect people with the wonders of the arts and increase cultural literacy, inspire the next generation of artists, and serve as a catalyst for economic development in our region.

Your gift ensures that artists like Phoebe Apperson Hearst and Warren McCullough continue to provide opportunities for people to experience the arts for the first time, as well as introduce new audiences to the wonderful worlds of art and design that the Mattress Factory has to offer.

The results are far from perfect, but revising the copy would require substantially less time and effort for a development team than creating new material. This demonstration also shows some of the limits of text generation. The content it generates is entirely based on the content it is supplied. Programs like InferKit are quite good at elaborating on prompts generally, but the more content you ask it to create, the more inferences it must make. Eventually, the generated material becomes very repetitive, veers off topic, or starts completely making things up.

Where using AI gets tricky

Using AI to generate solicitation copy may be appealing, but it gets complicated when it comes to donor relations. The donor bill of rights, an electively used set of rights developed by the Association of Fundraising Professionals, the Association for Healthcare Philanthropy, the Council for Advancement and Support of Education, and the Giving Institute illuminates these issues well. Article 8 states that a donor has “the right to be informed whether those seeking donations are volunteers, employees of the organization, or hired solicitors.” This begs the question, to what extent does an AI need to contribute to the solicitation of a donation before a donor should be informed? Since AI generated text cannot be copy written, does a Nonprofit have any obligation to acknowledge an AI for authorship? While not illegal, these practices would almost certainly create a great deal of frustration or mistrust from a donor if they found out the heartfelt plea for a donation they received was generated by an entity that has no heart.

It should be stated that it is misconception to think of AI as being alive. It is, by its nature, adept at impersonating human thought processes, and while some in the field claim AI may be achieving low levels of sentience, most experts would argue we are not there yet.

Salesforce, the world's leading Customer Relationship Management platform, has made a public commitment to developing ethical AI. Part of this commitment has been the adoption of 5 principles to guide the development of their AI software.

  1. Being of benefit

  2. Human Value Alignment

  3. Open debate between science and policy

  4. Cooperation, trust, and transparency in systems and among AI community

  5. Safety and responsibility

While these considerations are commendable, moral use of AI will ultimately come down to how a user will engage the program.

So how can nonprofits utilize these technological advances without alienating their stakeholders? For starters, it would be best to avoid its use for individual solicitations. No matter how well intentioned your organization may be, you will never be able to undo the betrayal felt by any individual donor if their integrity of tier cultivation was compromised by them finding out they have been corresponding with a robot. Additionally, ensuring that AI is utilized to implement mission based goals in demonstrable ways may alleviate some skepticism from worried stakeholders.

Ultimately AI is like any other tool. While it may simulate human thought process, it is ultimately up to the operator to use it morally. Perhaps it is time to update the donors bill of rights to reflect the new AI bolstered workplace. Let’s see if InferKit can come up with something…

See this content in the original post

+ Resources

“Ai for Good: Principles I Believe In.” Salesforce.org, September 20, 2021. https://www.salesforce.org/blog/ai-good-principles-believe/.

Al-Sibai, Noor. “OpenAI Chief Scientist Says Advanced AI May Already Be Conscious.” Futurism. Futurism, February 10, 2022. https://futurism.com/the-byte/openai-already-sentient.

Bailey, Jason. “Ai Artist Gives ‘Perfect’ Ted Talk as Cyborg.” Artnome. Artnome, October 26, 2020. https://www.artnome.com/news/2018/11/15/ai-artist-gives-perfect-ted-talk-as-cyborg.

Burns, Ed, Nicole Laskowski, and Linda Tucci. “What Is Artificial Intelligence (AI)? Definition, Benefits and Use Cases.” SearchEnterpriseAI. TechTarget, February 23, 2022. https://www.techtarget.com/searchenterpriseai/definition/AI-Artificial-Intelligence.

DiResta, Renee. “Ai-Generated Text Is the Scariest Deepfake of All.” Wired. Conde Nast, July 31, 2020. https://www.wired.com/story/ai-generated-text-is-the-scariest-deepfake-of-all/.

“The Donor Bill of Rights.” Association of Fundraising Professionals, January 23, 2019. https://afpglobal.org/donor-bill-rights.

Gain, Vish. “OpenAI Opens Developer Access to GPT-3 Text Generator.” Silicon Republic, November 19, 2021. https://www.siliconrepublic.com/machines/openai-gpt-3-api-developers.

“How We Visualize the Impact of AI-Driven Donor Outreach.” Ruffalo Noel Levitz, December 3, 2021. https://www.ruffalonl.com/blog/fundraising/how-we-visualize-the-impact-of-ai-driven-donor-outreach/.

I, A. “A Robot Wrote This Entire Article. Are You Scared Yet, Human? | GPT-3.” The Guardian. Guardian News and Media, September 8, 2020. https://www.theguardian.com/commentisfree/2020/sep/08/robot-wrote-this-article-gpt-3.

Martel, Adam. “AI for Good: 3 Ways It's Going to Change the Donor Engagement Game.” Salesforce.org, December 2, 2020. https://www.salesforce.org/blog/ai-good-3-ways-going-change-donor-engagement-game/.

“Nonprofit Fundraising Statistics [Updated for 2022].” Double the Donation, January 13, 2022. https://doublethedonation.com/tips/matching-grant-resources/nonprofit-fundraising-statistics/.

Radford, Alec. “Better Language Models and Their Implications.” OpenAI. OpenAI, June 21, 2021. https://openai.com/blog/better-language-models/.

Rao, Seema. “Museums and Ai: Can You See the Future?” American Alliance of Museums, October 27, 2021. https://www.aam-us.org/2019/01/21/museums-and-ai-can-you-see-the-future/.

Rao, Seema. “Museums and Ai: Could Robots Be Your New Coworkers?” American Alliance of Museums, October 27, 2021. https://www.aam-us.org/2018/12/26/museums-and-ai-could-robots-be-your-new-coworkers/.

Rao, Seema. “Museums and AI: Is There a Ghost in the Machine?” American Alliance of Museums, October 27, 2021. https://www.aam-us.org/2019/02/06/museums-and-ai-is-there-a-ghost-in-the-machine/.

Reben, Alexander. “Five Dollars Can Save the Planet | Alexander ... - Youtube.” Youtube.com. TEDx Talks. Accessed March 3, 2022. https://www.youtube.com/watch?v=8QUq8JTvoNM.

Smith, Dave. “Gmail Can Now Autocomplete Entire Emails with a New Feature Called Smart Compose - Here's How to Turn It On.” Business Insider. Business Insider, May 16, 2018. https://www.businessinsider.com/gmail-smart-compose-turn-on-instructions-walkthrough-2018-5.

Team, The Donorly. “3 Ways AI Can Help Grow Your Donor Relationships.” Donorly. Donorly, July 28, 2021. https://donorly.com/thedonorlyblog/3-ways-ai-can-help-grow-your-donor-relationships.

Tue, Feb 22nd 2022 12:03pm - Mike Masnick. “US Copyright Office Gets It Right (Again): Ai-Generated Works Do Not Get a Copyright Monopoly.” Techdirt, February 26, 2022. https://www.techdirt.com/2022/02/22/us-copyright-office-gets-it-right-again-ai-generated-works-do-not-get-copyright-monopoly/.

Vincent, James. “OpenAI Has Published the Text-Generating AI It Said Was Too Dangerous to Share.” The Verge. The Verge, November 7, 2019. https://www.theverge.com/2019/11/7/20953040/openai-text-generation-ai-gpt-2-full-model-release-1-5b-parameters.