You’ve Been Augmented: 7 Tips for Working with AI

Whether you’ve noticed it or not, you’ve been augmented by artificial intelligence (AI). Algorithms are learning more about you and your purchase, communication, and travel behaviors every hour of the day (and downloading all night).

Algorithms and innovations in AI are an occasion for debate and discomfort among most major corporations. There are the initial obstacles of onboarding the new technology, changing old processes, getting everyone acquainted with how it works and then, of course, the dreaded responsibility that comes with relying heavily on something that may not be fully understood. Then there’s the fear of spending thousands of dollars on a workplace innovation and having people use it for a basic function (i.e. Alexa/Siri and using it exclusively as a timer).

Here are some helpful things to keep in mind while making the switch and to help embrace AI and algorithms as a valuable and dynamic part of your team.

1. Demand Transparency

Understanding how an algorithm works enables you to fully grasp its possibilities, and, perhaps more importantly, its limitations. Although the goal of a AI may be to help avoid obstacles, physical or otherwise, it is still bound by human limitations: hardware, software, etc.

This knowledge not only helps to ensure a better partnership with an algorithm, but it also helps to vet algorithms, and to determine which are poorly designed. Math guru Kathy O’Neil says the solution to solving for poorly made algorithms is transparency and measurement. She says researchers must examine cases where algorithms fail, paying special attention to the people they fail and what demographics are most negatively affected by them.

There are instances called “The Black Box” where we are aware of the inputs and outputs we plug into an algorithm, however, we don’t understand what the AI is doing to process those inputs into outputs. For example, Digital Medicine Researcher and Cardiologist Eric Topol discusses how humans are unable to predict gender based on retinal scans. That does not mean there is not a correlation, however. He says that when we feed an AI hundreds of thousands of data points consisting of retinal scans and gender identification, the AI is able to accurately determine gender based exclusively on the retinal scan with 98% accuracy. In this instance, we know the AI is correct, but have no idea how it got to the answer. AI continues to make full transparency more difficult as it learns at levels that we did not initially predict, and regularly communicates with us in ways that we do not understand. This growing divide in AI-to-Human communication is likely to be a point of tension in the coming years.

2. Set Boundaries

To optimize algorithm use, it is necessary to maintain a control variable. Testing and experimenting while using an algorithm means setting guidelines around which automation/algorithms you are using, as well as knowing where the algorithm stops and you begin. Not only will this help with project management and task management, but, when done well, will encapsulate step one and enable a more seamless and mistake-ridden workflow.

Algorithms have been in use for longer than you may think, in places you may not expect, i.e. The Justice System since the 1930s. These algorithms helped to make predictive analysis of perpetrators’ likelihood to commit a crime again in the future — today we are reckoning with this predictive analysis and it has come under scrutiny as stereotyping rather than logical prediction. Through processing the results of algorithms in real time while knowing the method and data pool, we are better able to use valuable human traits such as empathy, nuance, and understanding to better ensure a just and effective symbiotic augmentation.

3. Bend Time

Algorithms’ most evident augmentation is the ability to work at inhuman speeds, parsing through mountains of information in a split second. Use this. Understand what to delegate and how to best integrate the data efficiently into your workflow. Tools like Zapier can make the data come to you, rather than making you go to your data, saving an exponential amount of hassle over time.

4. Establish Trust

It is important that once you’ve implemented these first few tips, you begin to trust the algorithm so that you can grow together. From the initial Man vs. Machine chess match loss, to daily innovations in the medical profession, algorithms have been proven to work. Let them. Don’t carry a false hierarchy of the Man > Machine, capture the results and let them speak for themselves. However, this involves trusting yourself as well. Allow for the algorithms to empower you, rather than submitting to their seemingly omnipotent whims.

5. Learn Jeopardy

One of the most difficult parts of working with data is knowing what decisions to make with the data provided. Sometimes the problem at hand is not apparent until after taking a pH of your intended audience/culture — in this instance, like Jeopardy, the answer comes first and the question is discovered after the fact.

6. Promote Yourself To Supermanager

It’s possible that the new work climate has primed us all to have superjobs, and to supermanage them by combining a variety of traditionally separate skill sets into one master position. This role synthesizes skills from creative innovators, data-driven decision makers, and empathetic mentors and turns it into the perfect person to be at the end of an algorithm. Talk about an opportunity for PD…

7. Be Ready For Take Off

With an algorithm at your right hand, it’s possible that you may just hit exponential growth. Be prepared for it. Capture data along the way to allow your algorithm to improve, document your learnings of what went well, and amplify a message you stand behind.

Code that learns is a new development in the tech savvy workplace, and it’s here to stay. Implement them gracefully and intentionally to save time and augment your professional life.

For more on AI in education, see:

• Why Every High School Should Require an AI Course
• Let’s Talk About AI Ethics; We’re On a Deadline
• Getting Started with AI: Resources For You and Your Community

This is the second post in a series on Digital Discernment: a #FutureofWork Series on Teaching and Leading in the Age of AI. For other articles in the series, see:

• Deepfakes: Teaching Critical Consumption 

Stay in-the-know with innovations in learning by signing up for the weekly Smart Update.