Standard Linear Regression with Python

Regression OLS with Python
sns.regplot (x=”distance_Sqrt”, y=”price_Sqrt”, data=taiwan_real_estate, ).resid plt.show()# Plotting correlation heatmap import seaborn as sb import matplotlib.pyplot as mp dataplot = sb.heatmap(taiwan_RE_core_varr.corr(numeric_only=True), cmap=”YlGnBu”, annot=True) # Displaying heatmap mp.show()

AI is changing everything

human and robot fingers touching each other

We have had AI around for more than a decade now, but for most people, there was nothing earth-shattering about it. As in, humans did not see their world change in a dramatic way just because there was a chatbot on their bank’s website, or their email client sorted their incoming messages into folders. Maybe because that felt more like automation than actual, capital letter Intelligence.

Capital letter Intelligence is something that only humans have, at least that is what we commonly thought. At the same time as AI has made more and more progress, we rethought our understanding of intelligence when we learned about how octopi perceive their world. The movie My Octopus Teacher arrived just in time – was that a coincidence?

In a podcast with 42courses, Chris Duffey, a leading tech futurist said that AI has been becoming more emancipated (my word).

AI, at the at the starting point, … about 7 to 5 years ago it was around this notion of being an assistant. It’s somewhat evolved to a more active role in terms of copilot.And now a number of AI companies are exploring this notion of almost an executive coach where an assistant or copilot feels like it’s still doing things that the user is asking it to do without proactive measures …

This evolved, emancipated AI was inevitable based on the culture of information technology players. Once a ball like this gets rolling, many want to take a turn at kicking it further.

A standard followup question is whether that is a good thing. Is that progress? But what I think is even more important to ask, and is often ignored, whether it is worth the very high price we and future generations are paying,

A very high price in terms of cheapening creative and intellectual work, and the high energy AI devours. I agree with Chris Duffey that what AI’s capabilities are phenomenal, and mindblowing, like a circus trick. It is amazing that some people can be transported by driverless cars, and that novels and poetry written by AI can win literary awards, that we can just use AI to compose music for our videos, etc. There is much to be said for flattening skills.

But what about the humans? What are they gonna do? What will bring them joy and fulfillment? and what will earn them a living? What will motivate them to develop skills and sensibilities, not to mention such virtues as persistence? What will the masses of humans do with their lives?

Principles of Display Design

cluttered controls in the three-mile-island nuclear reactor control room

Most B2B software includes displays of data, including dashboards and charts. While many of them are purely informational, the most important ones allow its users to interact with the data, or trigger some actions (either within the platform or outside of it). Just to give you some examples of possible actions:

  • For a retail business, triggered actions include ordering more of a certain product, cancelling orders, creating invoice, etc.
  • For a professional training company, HR professionals who monitor training compliance data may see the need to encourage or congratulate an employee depending on their success at completing the assigned training. Many of the managers and HR people prefer to do that in person instead of relying on impersonal emails that admin site allows them to generate.

Some of the display design can mean a matter of life and death as in the now infamous Chernobyl control room and the Three Mile Island Nuclear Power Plant, two places where the data display’s design was far from being human centered. Flawed display design cost untold number of human lives and a far reaching environmental disaster following the disaster in April of 1986 in Ukraine, and in Pennsylvania in 1979. I lived relatively close to Chernobyl (in Hungary) at the time of the explosion, but only decades later did I come to realize how easily that tragedy could have been avoided. Risk-tolerant tourists now can book a tour to visit the site in Ukraine and marvel at the dysfunctional design.

Designing data displays requires a deep understanding of the system, the admin’s needs for sources of information, and the risks that go with not having access to the right data at the right time. A tall order for a UX professional unless they can collaborate with an SME.

But there are some general principles to help.

N Principles of Display Design

Based on Attention

1. Make salience compatible with importance

Some pieces of information are more important than others, in fact, there is a whole hierarchy in information architecture. Accordingly, the most important information needs to be the most salient, and the less important pieces less salient. Otherwise, users get confused and or overwhelmed with the abundance of alerts competing for their attention.

2. Minimize information access cost

Information access cost is the time and effort it takes to locate the right page / graph / table needed. Good design makes sure that cost of traveling between frequently accessed sources is small. A related principle is

3. Display Mental Proximity

If the user needs to integrate two or more sources of information in order to complete a task, it is crucial that the user perceives those sources as clearly connected to each other (mental proximity) as possible. When we look at a line graph, we need to integrate the ups and downs of the line with the legend and the values. This is often achieved by having the various elements close to each other (proximity), but there are other ways to make the connection salient, such as using the same color and/or format, adding a line, etc. The reverse applies as well – avoid using the same color, pattern if you want to separate elements.

The line graph is in close proximity with the legend and the values. Having the same color for the graph and the legend further increases the connection.

4. Avoid resource competition

If you need to communicate multiple unrelated pieces of information to the user, use different modalities. For example, speed of a car can be seen visually on a dashboard, but the alert about seat belt is auditory (which can get very annoying).

5. Make the assets easy to discern and discriminate

When looking at a lot of information (e.g., dashboards, charts, maps), users do not want to engage in a “spot the difference” game. In other words, consumers of displays can get easily overwhelmed by minute differences as they demand a lot of cognitive resources.

Color One common example is a map with hues of one or two colors, which makes discerning the differences a challenge.

Text Hard-to-read, not legible text makes scanning very difficult. One way around it the Tallman method, which highlights difference
in drug names by capitalizing discriminating letters and making them bold is an example of making labels more. This is one of the techniques the FDA approved to avoid look-alike and sound-alike (LASA) drug names. (source)

FDA Name Differentiation Project

Similarity causes confusion. Similarity is the ratio of similar features to different features. Thus, AJB648 is more similar to AJB658 than 48 is similar to 58, even though in both cases only a single digit is different. Where confusion could be serious, the designer … should highlight dissimilar ones in order to make them distinctive.

Lee and Wickens

Sources consulted:

  • Wickens, C. D., & Liu, Y. (2017). Designing for people: An introduction to human factors engineering. Createspace Independent Publishing Platform.
  • https://www.fda.gov/drugs/medication-errors-related-cder-regulated-drug-products/fda-name-differentiation-project, retrieved on 23/2/2024

You have to wonder …

… what kind of research went into a probably expensive decision to build and set up vending machines for selling bitcoin in a grocery store. I am just thinking here of the user journey mapping, or maybe persona building based on shoppers frequenting Cub Foods locations (for those not familiar with this relatively low-priced grocery store chain). But in all honesty, I just can’t wrap my mind around it.

Is there a segment of customers who is comfortable investing their money in a public place (this was right next to the exit of the store) – so they do not have access to internet anywhere else, and at the same time into bitcoin?

Defining research questions

A basic expectation from UX researchers is that once they are equipped with the research questions, they talk to a certain number of consumers/ participants / users, analyze the qualitative and quantitative data, and then deliver the answers. Analyzing the data is a somewhat subjective, interpretative activity: there are no objective truths. Instead, as a researcher, you develop an understanding of your users’ needs, pain/ delight points, attitudes, and desires. You often discover something valuable that nobody has explicitly said. Like a prospector, you find the gold nuggets hidden behind the explicit verbiage. There are several frameworks for this process (grounded theory, thematical analysis, etc.) and it is understood to be the bread and butter of our work.

Less recognized is the process of defining the relevant research questions. This is where you need to look at your internal stakeholders (PMs, Designers, Content strategists, etc.) as your first group of users. You need to put in the same gold nugget prospecting and mining work as you will with your external users. In other words, you need to be as attentive to your stakeholders’ words as you are with your external users, and find the gold nugget questions behind the verbiage.

Erika Hall’s approach provides a good framework for that.

  1. Invite your stakeholders to brainstorm questions on a regular basis. Instead of the usual idea dump, participants are invited to contribute questions that stem from either “a lack of clarity or a wellspring of curiosity”.

2. Then, as a group, discuss which questions are high priority versus less risky to let them go unanswered, and which questions are the stakeholders the most in the dark about. This enables the team members to have a shared understanding and also to prioritize research.

3. Rinse and repeat. When done regularly, team members get used to the idea that they are not expected to have all the answers. It is OK to not know something and have the humility and curiosity to seek answers.