Nonprofits and ChatGPT…buds forever?

Welcome to data uncollected, a newsletter designed to enable nonprofits to listen, think, reflect, and talk about data we missed and are yet to collect. In this newsletter, we will talk about everything the raw data is capable of – from simple strategies of building equity into research+analytics processes to how we can make a better community through purpose-driven analysis.

Today you and I are talking about something pretty, shiny, too friendly, probably concerning, and possibly inevitable addition to our future – ChatGPT.

Imagine you met a stranger at a conference who seems interesting, fun, and easy to talk to. You don’t know much about their background, but they seem to be friendly enough. You pick any topic, and they come back with a witty response. They can hold an engaging conversation. They offer you their contact information, so you can connect any time of the day to chat. Consider yourself best buds already?

Not the perfect comparison, but ChatGPT is somewhat like that – a chatbot minus the human element. If Quora has helpful/not questions for you from different corners of the world, ChatGPT has useful/not answers for you.

Based on OpenAI’s language model GPT-3 (GPT stands for Generative Pre-Training Transformer), ChatGPT is a chatbot - designed to communicate with humans in a natural language. The model behind this chatbot is trained on texts from the internet - to create grammatically and syntactically articulate responses.?

If you haven’t already, Google ChatGPT, and a page like the one below will come up. You can type your usual questions and see how promptly it responds.

Unsurprisingly, the internet became obsessed with it in the past 90 days (since its launch late last year).?

ChatGPT seems to be helpful in more ways than one, from content creation to reducing time on everyday tasks. But there are a few points to consider before we can pledge our allegiance to this technology. The model behind ChatGPT requires training on a ton of data and powerful computational resources to deploy successfully for public use. Access to resources to that level is within reach of few. That means the friend we made at the conference in our example above – we are not going to know their background. In terms of technology claiming to be part of daily life, that’s half knowledge. And half-knowledge can be dangerous.

Acknowledging that ChatGPT can be both advantageous and disadvantageous to us in the long run, here are three lists for you and me to consider in the context of nonprofits. Specifically,?

• 5-ways when ChatGPT can support nonprofits,
• 5-ways when ChatGPT can cause harm, and,
• 5-things we can do to work better with ChatGPT

?

I. 5-ways when ChatGPT can support nonprofits

ChatGPT can give a nonprofit

1.ideas for creating social media posts.

E.g., “give me ideas for social-media posts around immigration and refugee needs.”

2. support in drafting letters of invitation (e.g., events, galas, etc.)

E.g., “write a letter of invitation for my annual gala to my lapsed donors.”

3. content for the website and micro-site.

E.g., “write a paragraph on why we must fight for immigration rights.”

4. ideas for group activities at the next team meeting.

E.g., “give me ideas for good icebreakers at the next team call.”

5. ideas of external audience engagement

E.g., “give me ideas for engaging my volunteers and external meetup group supporters.”

?

II. 5-ways when ChatGPT can cause harm

1. Lack of source information:?Without any source of its text output, you can neither be sure to trust it nor attribute a note of gratitude to the source of the response.
2. Lack of transparency on underlying data’s privacy:?Without the ability to attribute responses to identifiable indicators (like the author, date of creation, medium of content creation, etc.), we risk clarity on ownership and data privacy.
3. Biases in the underlying data:?The data training model of the chatbot is unquestionably coming with its biases. It can cause perpetuating stereotypes and representational biases in the responses. That means the responses generated need double-checking (depending on the question asked) – which can be a challenge for the non-native language speaker using ChatGPT.
4. Confirmation bias:?If you ask ChatGPT a question like “why are donations in nonprofits calculated as…”, it immediately starts listing four to five reasons as a response to the why. It does not challenge the question, provide differing responses or, at the least, ask for more context before generating a response. By?agreeing to most questions, ChatGPT allows confirmation biases. Confirmation bias, here, means the user can interpret ChatGPT's response to be positively consistent with the underlying belief of their question – thus, perpetuating bias.
5. Lack of transparency behind data sources and methodology:?the lack of transparency behind sources of data and methodology can lead to over- or under-representation from communities. That means the users can’t be sure if the response comes from the representation of a particular location or identity.

III. 5-things we can do to work better with ChatGPT

?Now, what do we need with ChatGPT that can make its use more robust for us?

Here are 5-things we can do at this moment:

1. Choose contexts?carefully:?Choose contexts of ChatGPT use wisely. There is a difference between asking, “what fun topics can I write about donor engagement on my social media?” vs. “write an onboarding email to my new board member.” Remember, regardless of your chosen context, you must verify the output text before its use.
2. Continuous education:?Invest in continuous education for yourself and your team so that you collectively learn what engaging with any algorithm (not just ChatGPT) should mean for you.
3. Collective feedback collection:?Build a process in your nonprofit for collecting feedback from the texts produced by this chatbot. That way, shared accountability can make better use of this technology.
4. Remember consequences:?Using text-output of ChatGPT comes with consequences. That means, before producing, sharing, or using content from the technology, check if the text output is human-centric, not perpetuating direct harm to individuals or a specific community.
5. Invite, encourage, and embrace diversity:?If AI (specifically generative AI like ChatGPT) is inevitable for our future, it is imperative to embrace diversity in our workplace cultures. We need people of all identities and backgrounds – who come with the knowledge of math, science, social science, civics, and economics – to make this technology human-centric for us. With meaningful diversity, we can hope to understand such technologies through a technical, legal, ethical, social, racial, and political lens.

*********************************

We like it or not, more AI in the future is inevitable. Big tech companies wrapping technology in shiny wrappers for more followers amongst us – yes, that’s also a highly likely possibility. Does this mean the technology needs to be tagged as our frenemy immediately? Nope. We are the data being fed into the algorithms, and we are also the consumers of its output. So we don't have to pick lanes with new technologies immediately.

This is our chance – yours and mine – to consciously learn together as social-minded human beings before we make a claim on the future. The world can be notorious for making us forget one of the two -of our rights or responsibilities - at a given moment in time. It is upon us to remember that we have both and that we need to exercise both for an inequitable, healthy, meaningful planet and life.

AI has a future?because of us,?with us.