If you were going to ask the person next to you for money, what would you want to know about them? According to David Woodruff of MIT and Rich Palmer of Gravyty, this is a vital question in a 400 billion dollar industry. When you think of all of the embedded ways we give, it’s a massive part of our culture scattered across a vast number of models, industries, and methods.
The application of nonprofit fundraising is a big question that both want to answer. In their talk for ODSC East 2019, “Expanding Nonprofit Workforce with Deep Learning,” they take us through the primary system the partnership has created. It involves the application of data to a potentially sensitive sector: philanthropy.
As the amount of money grows, the necessary information grows in tandem. There’s a big difference between asking for something like $10 and asking for $1 million:
- Do they have the assets?
- What is the story that will help them decide to give?
- What is their passion?
Finding that data is part of formulating a strategy for successful donations. Whether you belong to a small nonprofit or a large one with millions of potential dollars on the table, systematizing your strategy could help create a more reliable pipeline for these major gifts.
Setting the Stage
The state of philanthropy involves three separate areas:
- emotional intelligence
You must be able to have a trusted conversation with donors in order to be successful. This is a direct link. The ability to raise major gifts is part of this. 95% of your money raised comes from just 5% of your market, so understanding this psychology is vital to the success of the organization.
A second layer involves getting to those unrated donors to create a broader pipeline of donors to give major gifts. According to Woodruff, it takes about 33 months to get to that major gift, so you must be able to understand and optimize this large unrated pool.
The lack of data isn’t the issue. Talent hasn’t quite saturated the nonprofit world yet, so it’s data left on the table. Second, there’s a specialized talent involved in fundraising that doesn’t include anything like school.
The technology could help fill this gap faster than cultivating human talent alone. Once a nonprofit figures out the pool of unrated potential donors, the quest for data is the question.
Open Source Methods for Language Data
Gravyty is using proprietary data, but other open source language models can shed light on how to quantify (and therefore qualify) this language data:
- SpaCy: part of speech tags
- Scikit-Learn: Timing algorithms (random forest0
- TensorFlow: Platform to generate models
- NLTK: tokenizing sentences and words
When you apply AI to the philanthropy sector, you can do things like apply these language models of that of an email frame to learn how the most accomplished fundraisers both phrase and time their communications with potential donors.
Using these models allow people to have more meaningful conversations with their potential donors. It streamlines the process and raises more money with fewer people. In a field in which labor comes at a premium, this is a big deal.
Applied AI and the Future of the Workforce and Deep Learning
The group has formed an advisory council to kickstart AI in the nonprofit industry. The council advises on both practical as well as ethical issues. It’s crucial for us to understand the delicate issues surrounding trust and giving and how the intrusion of AI could potentially impact people’s giving.
AI is seeing a rebirth in interest, and the work of MIT and the council provides a way forward for the nonprofit sector and AI. These conversations can help advance fundraising and ease pressure on nonprofit groups. The system isn’t meant to replace the workforce but to provide tools to augment the labor of those working in the nonprofit industry.
The nonprofit sector is often behind in adopting new technology. Understanding how people will interact with this technology for the first time is another critical piece of using AI to improve relationships between nonprofits and their donors.
Gravyty sought to work with the ways that people are already working by expanding the workforce with deep learning. Despite the difficulty of working with advisory councils, the company wanted to get ahead of the potential issues presented with using this type of technology in the unique environment of nonprofits.
Their talk on the workforce with deep learning focused on the basic needs of nonprofits for both labor to do the discovery phase and the tools to understand the ways that some of the best fundraisers in the field are working. Using these open source methods, nonprofits could find themselves finally adopting the technology the business world has embraced.