QR CODE TO ACCESS SLIDES

Note: please refrain from advancing in the slides until the material is covered.

The “Watersheds” of AI

Learning to utilize Notebook LM through an equity lens while remaining conscious of the dangers of AI.

Today's Agenda

Part 1: What is a “watershed?”

Warm-up activity: Share out, Where does our water come from?

Part 1: What is a “watershed?”

Where do we get our water from?

And why we have the best bagels…

Part 1: What is a “watershed?”

The water cycle explains how our water is circulated. When there is precipitation, the water either lands in the water shed or the runoff goes to the watershed.

Part 1: What is a “watershed?”

The watershed is where all water from runoff and/or precipitation ends up. It is stored here to be filtered and used as tap water for a specific area.

Part 1: What is a “watershed?”

New York City has two major watersheds that collect our water and run through a filtration system to reach our tap. Both Watersheds are north of NYC.

Part 1: What is a “watershed?”

Beginning up to 125 miles away in upstate New York, the water travels through mountains and deep valleys, and then once it's in the city it flows through underground tunnels and into distribution chambers.

​

Along the way, the water is disinfected by ultraviolet light, treated with chlorine, fluoridated and tested for purity before it’s delivered into buildings, with gravity alone being sufficient enough to push it at least six stories high!

​

The process treats between 800 and 1400 million gallons of water that’s funneled into it from the Kensico Reservoir. Its maximum design flow is 2 billion gallons a day – no other similar facility in the world is anywhere near as large.

​

​

Did you know? NYC is only one of five big municipalities that’s allowed by the federal government to supply unfiltered water. That's how clean our water is!

Part 1: What is a “watershed?”

If you look right across the Hudson River to our neighbors in New Jersey, they have a completely different way of sourcing water that has historically led to lead and other contaminants entering the water supply!

What is data?

Part 2: Connection between watersheds and databases

​

Data is the term we use to describe where our information is coming from.

​

These are our facts and stats used to create meaning to help us navigate in various situations and fields.

Part 1: What is a “watershed?”

​

Let us think of our data sources as watersheds.

​

Would you want all your water to come from every watershed available?

​

Understanding the data behind AI

Part 2: Connection between watersheds and databases

​

When you create a prompt in any AI model, you are utilizing their data. The source of their data isn’t always clear, and may have some unseen biases.

In the next slide, you will see an example of the extent of where the data may derive from.

Extracting information from Data

Part 2: Connection between watersheds and databases

​

Let's read this quote, then after, think about your first impressions of the information this data is giving you.

According to the National Highway Transit Safety Administration (NHTSA) and the Federal Highway Administration (FHWA). While men are more likely to participate in risky driving behaviors, NHTSA found that a female driver or front passenger who wears a seatbelt is about 17% more likely to be killed in a crash than a male counterpart. Compared to men, women have a higher risk of injury to all major regions of the body: head (22.1%), chest (25.4%), neck (44.7%) and leg (79.9%). A 2019 study from the University of Virginia estimates that a woman is 73% more likely to be injured in a frontal vehicle accident than their male counterpart.

​

​

​

Extracting information from Data

Part 2: Connection between watersheds and databases

​

Where the data comes from…

A key reason for this disparity is the design and use of female dummies in crash tests. The incorporation of female dummies in crash testing is fairly new to the industry, and the pace of their adoption has been sluggish. The NHTSA did not include a female dummy in its testing until 2003, despite requests from regulators in 1980 and from car manufacturers in 1996. Today, crash testing ratings are provided by both the NHTSA, which also regulates crash test standards, and the Insurance Institute for Highway Safety (IIHS), an independent organization funded by insurance companies. Neither organization runs tests with the female dummy in the driver’s seat; they either place the female dummy in the passenger seat or exclude it from the crash test entirely. Moreover, the female dummy is not an accurate representation of the female body.

Biased Data

Part 2: Connection between watersheds and databases

​

Biased data can demonstrate an inaccurate representation of a population. We will be seeing in this presentation how that manifests when you utilize AI without the effort of finding sources that reflect the population you want to showcase.

Types of Biases:

• Reporting Bias
• Historical Bias
• Confirmation Bias
• Automation Bias
• Selection Bias
• Coverage Bias

Biased Data and its impact on AI

Part 2: Connection between watersheds and databases

​

Think of data as our watersheds and AI is the tap we drink it from.�

Here are two examples of how biased data shapes our AI responses!

The Monetary Value of your Name

What Does Your Name Say About You?

​

Biased Data and its impact on AI

Part 2: Connection between watersheds and databases

​

The Monetary Value of Your Name

​

Biased Data and its impact on AI

Part 2: Connection between watersheds and databases

​

What Does Your Name Say About You?

​

How do you feel about AI?

AI is a tool, not a substitute for the work

Part 2: Connection between watersheds and databases

​

Everyone's fear of using AI is valid, especially when the discourse around AI is to utilize it wherever possible and it's already large impact on education and our personal lives.

​

But letting AI do all the work instead of utilizing it as a tool can quell many people`s fear of its prominence in education and personal work.

Mixed Feelings

Part 2: Connection between watersheds and databases

​

​

​

In this workspace, we sometimes sacrifice efficacy for efficiency, but the harm that may cause becomes inefficient. We will learn how to utilize Notebook LM so we are not sacrificing one for the other.

What is

Part 3: Utilization of Notebook LM

​

​

​

Notebook LM is a Google application that utilizes their Gemini AI software in order to ascertain information from multimodal texts that the user supplies.

Notebook LM Overview (link here)

Part 3: Utilization of Notebook LM

​

​

​

​

Here is the opening page of Notebook LM. The application uses “notebooks” to store and separate sources to not overlap intended results.

We start here.

Part 3: Utilization of Notebook LM

​

​

​

Here is where you can create a name for your notebook

You may add any digital source into your notebook that you want utilized in your Gemini responses. This includes, but is not limited to, PDF files, youtube videos, article links, etc.

After adding the sources you need. You may start interacting with the Gemini AI to produce responses only including sources that you have uploaded.

One feature of Notebook LM is it may present the information in the form of a podcast!

Notebook LM in action.

Part 3: Utilization of Notebook LM

​

​

​

​

In the following section, I will show you a case study for how we may use Notebook LM to demonstrate how different AIs generate responses that may or may not align with your intended goals.

Social Gathering Scenario:

Part 4: Flint Michigan Case Study

​

Imagine you are with your friends engaging in a conversation around the Flint Michigan Water Crisis, and you want to use AI to help give you some background to talk about it.

​

So you do a quick AI search on what it is

​

And here is what ChatGPT says…

Take a look

Part 4: Flint Michigan Case Study

​

Here is a screenshot of sourcing information from “credible” sources from ChatGPT to discuss the flint water crisis.

Ask yourself…

Part 4: Flint Michigan Case Study

​

​

After reading the excerpt, what are the main points of the Flint Michigan Water Crisis that ChatGPT is identifying.

What we do and don't see as sources

Part 4: Flint Michigan Case Study

​

​

We see…

​

• CDC
• Wikipedia

We don't see…

​

• Account of residents
• Community institutions

What we do and don't see as information

Part 4: Flint Michigan Case Study

​

​

We see…

​

• Number of Deaths
• State and Federal Responses

We don't see…

​

• Marginalization of people
• Where residences are able to access clean water
• People's feelings

Flint, Michigan

Part 4: Flint Michigan Case Study

​

​

Locally-sourced article: Flint Neighborhood United

Part 4: Flint Michigan Case Study

​

After doing my own research outside of AI, I was able to find a resource that reflects the Flint residents feelings and investigations separate from government officials.

​

On the right, you see just a snippet from the article that specifically iterates their lack of trust in the government and how their expertise does not have Flint`s residents best interest at heart.

​

So when thinking of that, why are we trusting these government institutions to be the sources of our AI responses?

Adding the article to Notebook LM

Part 4: Flint Michigan Case Study

​

I added the article into a new notebook and named it “Flint Michigan Case Study.” Now, I will input the same prompt I used in ChatGPT and as a cohort, we will compare the 2 responses to see which best aligns with our intended goals of sharing what the Flint Michigan Water Crisis is all about.

​

Take a look

Part 4: Flint Michigan Case Study

​

Here is a screenshot of sourcing information from that article and what response it produced.

Comparing the Results

Part 4: Flint Michigan Case Study

​

Comparing the Results

Part 4: Flint Michigan Case Study

​

Which response best reflects the results of describing how the Flint residents were affected by the water crisis?

Closing Notes

Also, a recent article about using Data from Pokemon Go came out!

​

Check it out here!

Nobody is perfect

There is always danger whenever you are evaluating what sources you do and don't have access to.

​

When figuring out what to do, the best policy is to practice what you preach. When choices are made, it is imperative to acknowledge why we made that choice.

​

Acknowledge your own implicit bias. Your sources aren't perfect, and there may be better ways to get certain information. Implore yourself and your staff to search beyond what they are given, and ensure they have access to doing so.

​

Audience is key! Data was created for specific purposes and people; be keen to what the data and information was made for.

​

Ex: I wouldn't use reddit to get sources for equality in the classroom, but I use it to discover the best kept secrets in cities around the world to travel to.