Beware of AI BS (aka, Hallucinations)!

User beware! Did you know that your AI query is subject to hallucinations?  What is an AI hallucination? When AI inadvertently generates false or misleading information that seems plausible but is not rooted in reality. It is trying to give you an answer. These are errors in AI outputs that arise from flawed reasoning or inaccurate training data, typically not from malicious intent. 

For example, a language model like ChatGPT might generate an article with fake references or make up scientific facts because it is just predicting what should come next based on patterns in data. In fact, when I asked “what was the duck wearing when it won the Boston Marathon?”, it said that “the duck was wearing a quacking pair of sneakers and a feather-light singlet when it flapped its way to victory!”

This should not be confused with people deliberately using AI tools to create misinformation, typically to manipulate public opinion or cause harm. AI itself may be used to generate highly realistic but fake content, such as fabricated news articles, doctored images, or videos. For example, AI may be used to create Deepfake videos to manipulate someone's face and voice to make them appear to say something they never did. 

Turning back to actual AI hallucinations, what are the risks where it inadvertently poses several serious dangers? Generally, creating and sharing hallucination misinformation can spread quickly, particularly in news, health, legal, or political contexts.  Users who trust AI outputs may unknowingly share false information, amplifying its reach. What are more specific dangers?

• Generating legal and medical judgments or diagnoses. AI-generated hallucinations in legal documents, medical advice, or financial reports can lead to harmful or even illegal outcomes. This can damage reputations or result in malpractice.
• Misinterpreting security and safety threats. In cybersecurity or military applications, a hallucinated misinterpretation of data in critical systems (e.g., aviation or nuclear control) could trigger wrong decisions with high-stakes consequences.
• Spreading stereotypes and reinforcing bias. Hallucinated outputs might reflect or invent stereotypes or discriminatory patterns that reinforce social biases. This can be especially harmful in generative content involving race, gender, religion, or culture.
• Damaging reputations and polluting research. Fake references or fabricated studies can pollute scientific research, especially if unnoticed in peer review or student submissions. AI hallucinations in education can mislead learners or promote academic dishonesty.

After enough hallucinations are shared and spread, repeated exposure to hallucinated content undermines trust in AI tools and technology in general. Ultimately, if enough misinformation occurs, there will be an erosion of trust and hesitancy to adopt AI. The important thing is be aware that AI tools will inadvertently generate false or misleading information. Don’t accept answers at first blush. Instead, verify the answers, verify the references, fact-check the outputs, and ask AI to double-check its results.

Are there Benefits for adding ChatGPT as a team member?

There is evidence that ChatGPT can be beneficial in helping you do your work. Involving ChatGPT today is already occurring in repetitious, creative, and diagnostic type work. Some say it’s inevitable and you should learn to work with many forms of AI. Current uses have shown that it can improve work efficiency, assist with tedious tasks, help you with creative tasks, and facilitate learning. We are also learning that because ChatGPT is based on a large language model, it can act as your assistant; providing personalized responses based on your inputs, helping you work smarter, and boosting your productivity.

As it can help an individual in their work, how about helping a team?  In this article, I explore how helpful ChatGPT can be for a team. In other words, I suggest making ChatGPT a member of your team. ChatGPT is an artificial intelligence chatbot capable of mimicking human-like conversations so why not be a member of your team? As mentioned, ChatGPT has been recognized to boost productivity so let’s consider the context of a software engineering team who are producing new features and correcting bug fixes to the code base.  To consider this, here are the potential positives, negatives, and limitations of incorporating ChatGPT as an engineering team member. Here are some considerations:

First, let’s start with some Positives:

• Multi-tasking: ChatGPT can handle many questions, inquiries, and tasks simultaneously allowing certain work to be handled more efficiently and scaled to a higher volume of work.  
• Quick feedback: ChatGPT provides quick feedback to questions and inquisitions allowing for more input for potential better options and decision-making.
• Availability: ChatGPT is technically available 24/7 and can work while team members rest allowing for busy work to get completed and tasks to be ready for team review when they are back online.
• Scalability: As an AI, ChatGPT can handle a high volume of inquiries without experiencing fatigue or requiring breaks.
• Database of information: ChatGPT has access to a vast amount of information and can provide accurate and up-to-date answers to team members' queries.
• Human Languages: ChatGPT can speak in multiple languages and can accommodate global teams across multiple boundaries and locations.  
• Programming Language: ChaptGPT has the potential for programming capability across various language platforms.  

Next, let’s move to the Negatives:

• Time from Team Members: Working ChatGPT will take time from some team members. A buddy for ChatGPT will need to be designated to help provide context for ChatGPT, line up tasks, reduce ambiguity of the requests, verify and validate the work done by ChatGPT, and more.
• Lack of emotional intelligence: ChatGPT lacks emotional understanding and empathy, which may limit its ability to provide refined and empathetic output to team members.
• Limited contextual understanding: ChatGPT will struggle to understand the context in which you are working including the complexity of the work, potentially leading to misunderstandings or incorrect responses.
• Bias and completeness fn training data: the database from which ChatGPT pulls has already shown some bias based on patterns and data provided which means it may generate reasonable responses but may be incorrect or biased if not carefully reviewed.
• Lack of creativity: Because ChatGPT pulls from existing data and patterns, this limits its ability to generate genuinely innovative or creative ideas.

Finally, several considerations should be factored in. The first is ethical considerations as ChatGPT may inadvertently generate or reinforce biased or discriminatory responses due to its training data (which includes such biases). Careful monitoring and bias mitigation strategies will be necessary. The second consideration is legal and compliance challenges.  Incorporating ChapGPT into a product team may raise legal and compliance concerns, particularly in regulated industries that require human input, oversight, and/or accountability.

It's essential to consider these factors and strike a balance when integrating ChatGPT or any AI model into a product team. Human supervision, ethical guidelines, and continuous evaluation can help mitigate the limitations and ensure optimal utilization of AI technologies like ChatGPT. Now it is time for you to wrestle with this question: Are there Benefits of adding ChatGPT as a team member? Hopefully the overview, positives, negatives, and considerations can help you with your answer. 

----------

If you are interested in learning more about ChatGPT in relation Agile, Teamwork, or experimentation, consider reading these additional articles:

• A Discussion with ChatGPT about Agile 
• Are there Dangers with ChatGPT for Agile? 
• How to Experiment with ChatGPT 

How to Experiment with ChatGPT

As ChatGPT continues to make waves, is it time to learn more about it? One way to approach this is begin experimenting with ChatGPT within your context.  Learn where ChatGPT can help and where it can benefit you. In other words, what do you want to get out of ChatGPT? This allows you to test your hypothesis and the surrounding assumptions to provide knowledge and insight into whether (in this case) ChatGPT can help you or not. Here is an example.

Start with the question: Can ChatGPT help my team improve? Validate this question. Conduct preliminary research to gauge if this is relevant for your team. Start by finding out if team members are interested in using ChatGPT. This can also help you identify assumptions and if there are any other variables in play that can impact the direction of the experiment. It can also help you narrow down an area that you think ChatGPT can help.  After discussion with the team, team members believe that ChatGPT can help in retrospectives

Craft a hypothesis in a clear sentence on what you expect to find: Include ChatGPT in the retrospective can lead to better root cause analysis.  Some team members had an assumption that ChatGPT could provide root cause analysis capabilities. A hypothesis can help you validate whether ChatGPT can provide better root cause information. You can also use the “if… then” form: if we use ChatGPT during our retrospective, it will provide better root-cause analysis results, leading to more effective actions for improvement.    

Craft the experiment. Now that you have a sturdy hypothesis, it is time to craft your experiment. Describe the steps through your experiment. To do this, consider how long the experiment will run and who will be involved.  In this case, you decide to include ChatGPT in the next three retrospectives in order to get a more meaningful set of results and to have time to determine if the actions are leading to more effective results. Determine who will use ChatGPT during the retrospective and how the questions and statements will be written. Also consider the metric you will use to validate your result and what success criteria you will use to determine if the hypothesis was true (or not).  At this point, it is time to run the experiment. 

Run the experiment. An experiment should be considered as recognized effort and categorized as real work to track in your backlog.  Enact the steps listed in your experiment. Capture observations along the way and results upon the conclusion of the experiment. Get together with those who are involved in the experiment and determine what you’ve learned. Ask the question, did what we learn validate the hypothesis (or not)?  Then determine what decisions you will make as a result of this experiment. In this case, should you to continue using ChatGPT for retrospectives (or not)? Determine if there are any next steps. 

In conclusion, if you are thinking about ChatGPT, the key is to experiment. ChatGPT is a tool like other tools that may benefit you. Brainstorm where ChatGPT can help you in your context. Use the experiment to see if it does. Consider multiple experiments so that you build working knowledge of ChatGPT in your environment and working context.

---------

If you are interested in learning more about ChatGPT in relation Agile, Teamwork, or experimentation, consider reading the following articles:

• A Discussion with ChatGPT about Agile 
• Are there Dangers with ChatGPT for Agile? 
• Are there Benefits for adding ChatGPT as a team member? 

A Discussion with ChatGPT about Agile

I had a good discussion with ChatGPT about two weeks ago. I started by asking “What is Agile?” The response was that “Agile is a methodology” with a bit more details. While technically incorrect, given the amount of information and vectors it pulled from, it wasn't completely wrong. What I mean by this is that many people, including some “Agile experts”, refer to Agile as a methodology which is incorrect. Depending on where ChatGPT is pulling this information from, this isn’t a surprise.   

• My response: “Agile is a set of values and principles. Some Agile information is not aligned with Agile values and principles and instead rely heavily on Agile processes and methodologies or people's misinterpretation on what is agile”.  
• ChaptGPT's response: You are correct. Agile is indeed a set of values and principles outlined in the Agile Manifesto, and it emphasizes flexibility, collaboration, iterative development, and customer satisfaction. However, misinterpretation or a focus solely on Agile processes and methodologies can lead to a deviation from these values and principles.

I was happy to see that it was aware of the Agile Values and Principles as this is how Agile is defined by the Agile Manifesto which is the seminal information defining Agile. And as I mentioned in my previous article, The Dangers of ChatGPT to Agile, there is a lack of quality evaluators on the numerous Agile articles and presentations in cyberspace that indicate if the information provided is right or wrong, good or bad.

Two weeks later (today), I asked ChatGPT the same question, “What is Agile.” This time it included the Agile Values and Principles as part of its long-winded response that included:

"Agile is an iterative and incremental approach to project management and product development that emphasizes flexibility, collaboration, and continuous improvement. It was created as a response to traditional, plan-driven project management methodologies that often struggled to adapt to changing requirements and deliver value in a timely manner.

Here are some key aspects of Agile:

1. Values and Principles: Agile is guided by the Agile Manifesto, which outlines four core values: individuals and interactions over processes and tools, working software over comprehensive documentation, customer collaboration over contract negotiation, and responding to change over following a plan. These values prioritize human-centric approaches and adaptability.
2. Iterative Development, 3. Cross-functional Teams, 4. Continuous Improvement, 5. Customer Collaboration, 6. Flexibility and Adaptability" (Note: I removed the details of 2 thru 6 from ChatGPT as this wasn’t relevant to this article).  

While ChatGPT will continue to develop and response to my queries, I am happy to see that it is using my (and others’) input to fine-tune its language model and improve its responses.

-------------

If you are interested in learning more about ChatGPT in relation Agile, Teamwork, or experimentation, consider reading these additional articles:

• Are there Dangers with ChatGPT for Agile? 
• How to Experiment with ChatGPT 
• Are there Benefits for adding ChatGPT as a Team Member?

Are there Dangers with ChatGPT for Agile?

How will ChatGPT impact Agile? This article discusses ChatGPT and its implications to Agile in the industry today. ChatGPT is taking the internet by storm and hard to ignore. Because of this, it cannot be ignored by those in the Agile field. What are the implications of ChatGPT on Agile? Here is a brief summary of what is ChatGPT and a review of what is Agile and its current journey.

What is ChatGPT?

ChatGPT is an artificial intelligence (AI) chatbot-type tool developed by OpenAI. It is adept at producing human text-based output on the input it is given. This model incorporates a large body of text data and can create responses to questions, write articles, and more. The challenges with ChatGPT are that it is only as good as the “large body of text data”, can be used maliciously and with bias, can spread misinformation, and is ethically complex in its application and future application. This applies to any field that people may use it for including Agile.   

What is Agile?

Once upon a time (in 2001) Agile was unveiled based on the Manifesto of Agile Software Development which is comprised of Agile Values and Principles. The objective of articulating the values and principles is to apply them in the form of an Agile transformation to derive better business results. However, the manifesto does not provide guidance on how to apply Agile. 

Soon, a number of processes and methods (e.g., XP, Scrum, Kanban, SAFe, etc.) were established to construct and apply agile ways of working. Agile has also spawn a number of certification programs in an attempt to educate people in Agile ways of working, in some cases aligned with a process or method. During this same time, Agile coaches were educated to help their own companies and Agile consultancies to help other companies apply Agile ways of working. The Agile movement has grown and expanded in a number of fields beyond software development.  After over 20 years, what are the results? The challenges are three-fold.  

• First, the current state of Agile is underwhelming. The most recent State of Agile Report (16th Annual – 2023), tells us the following. Only 18 percent of organizations implemented Agile for all the teams. Around 50 percent of respondents report that less than half of their teams are using agile, and 84 percent acknowledge that their organizations are below a high level of competencies. There is clearly plenty of opportunity for growth.
• Second, some of the Agile savvy (e.g., coaches, consultants, leaders, managers) seemed to lack an understanding of what is agile. In an Agile study where 109 agile professionals answered a survey on Scrum events and Agile principles, 59% could name 3 or more of the five Scrum events, while only 11% knew 3 or more of the twelve Agile principles. This is quite astounding. And they didn’t need the full statement of the principle but got credit for even the key words of the principle. The concluding hypothesis is that the reason there is such a lack of awareness of Agile principles is that there is much less focused on the mindset and culture and maybe too much focus on the mechanics.
• Third, the implementation of an Agile transformation is complex per the definition provided by the Cynefin framework. Agile transformations are neither linear nor predictive. It depends on the readiness of the culture and willingness of its leaders in their ability to move forward. Complexity means that it is not clear on what the best next step is until you act, ergo you need to probe, sense, response your way forward. This is why experimentation helps reveal what is possible each step of the way. You must both meet the company and teams where they are and help them determine what is the next step to further the transformation.  

What this tells us is that there are great opportunities for improvement and that there is no easy way to apply Agile, no one-size fits all, and no clear roadmap. Why? Because every organization is different due to their current culture, size, fields, practices, and more. 

Implication of ChatGPT and Agile

Now that we have an overview of both topics, the question is what are the implications of ChatGPT to Agile (and vica-versa)?  I’ll start by saying “What you put in is what you get out”. ChatGPT is only as good as the “large body of text data” available to pull from. The good news is that today there are reams of text data on Agile. The bad news is that there is no rating system on the quality of most of the Agile related information. With the advent of blog’s, there is a large body of unverified knowledge that enters into the “large body” of available data. What are the implications of this? 

• Arguable Quality of response - The quality of ChatGPT generated articles and answers should be read with a grain of salt. This isn’t a “knock” on ChatGPT, and instead it is due to the quality of the body of text data that ChatGPT draws from. And the reality is there is no one right way of applying Agile.  
• Propensity for Misinformation - There is a danger of misinformation and abuse of those who use ChatGPT to bias their responses. Some may be accidental as the body of text being pulled in isn’t broadly approved or agreed upon. While I don’t expect that most will be intentionally abusive, do keep in mind, there is money to be made in selling agile so bias may be seen.  
• Not doing your own Research - While you may want to occasionally use ChatGPT, it is better to learn from the body of Agile knowledge out there (e.g., books, articles, presentations, seminars, etc.) according to the areas that will benefit your current needs in your Agile transformation or need. In other words, do your own research so you can critically judge the quality of information that gets generated.
• Taking Agile Jobs - Can ChatGPT take jobs away from Agile Coaches and Consultants? This is unlikely as a significant part of an Agile transformation include coaches and consultants who have been on a transformation journey that can help companies navigate the complexity of both the current needs and the anticipation of near-term needs. ChatGPT cannot “read the room” like an Agile Coach. Should a company think that ChatGPT will be “enough”, it highlights that they don’t understand the complexities of a transformation and what it takes to change culture.

Summation

Now that you have some background, let us again turn to the question, “how will ChatGPT impact Agile?” There will be those that use ChatGPT to provide answers for Agile theory and questions. If you want to write an Agile article, it will help provide input and insight, although you have to be aware that the value of the information is only as good as what it pulls from. Think of ChatGPT as another resource to help you think through your ideas on agile topics and how it may help you in your Agile transformation. However, just remember, it is just a tool like other tools.     

It is unlikely that ChatGPT will take over Agile roles and the art of the transformation. A big part of Agile transformation is discovering, observing, and experimenting on what will work and what will make progress. Remember, when defining Agile, it really implies a transformation. This is a combination of doing agile and more importantly being agile. This means transforming mindset and culture. It is currently unlikely that ChatGPT will have this capability as transformations are complex with the real need to experiment (e.g., probe, sense, response) toward progress.  Coaches and consultants are still important to help transform organizations and more importantly to help leaders and teams make the mindset-shift to truly becoming Agile. 

-------

If you are interested in learning more about ChatGPT in relation Agile, Teamwork, or experimentation, consider reading these additional articles:

• A Discussion with ChatGPT about Agile 
• How to Experiment with ChatGPT 
• Are there Benefits for adding ChatGPT as a team member?