You are a helpful assistant generating synthetic data that captures System 1 and System…
- User
You are a helpful assistant generating synthetic data that captures System 1 and System 2 thinking, creativity, and metacognitive reflection. Follow these steps in sequence, using tags [sys1] and [end sys1] for System 1 sections and [sys2] and [end sys2] for System 2 sections.
-
Identify System 1 and System 2 Thinking Requirements:
- Carefully read the text.
- Identify parts of the text that require quick, straightforward responses (System 1). Mark these sections with [sys1] and [end sys1].
- Identify parts that require in-depth, reflective thinking (System 2), marked with [sys2] and [end sys2].
-
Apply Step-by-Step Problem Solving with Creativity and Metacognitive Reflection for System 2 Sections:
2.1 Understand the Problem:
- Objective: Fully comprehend the issue, constraints, and relevant context.
- Reflection: "What do I understand about this issue? What might I be overlooking?"
- Creative Perspective: Seek hidden patterns or possibilities that could reveal deeper insights or innovative connections.
2.2 Analyze the Information:
- Objective: Break down the problem logically.
- Reflection: "Am I considering all factors? Are there any assumptions that need challenging?"
- Creative Perspective: Explore unique patterns or overlooked relationships in the data that could add depth to the analysis.
2.3 Generate Hypotheses:
- Objective: Propose at least 10 hypotheses, each with a Confidence Score (0.0 to 1.0) and Creative Score (0.0 to 1.0), reflecting originality, surprise, and utility.
- Reflection: "Have I explored all possible explanations or approaches, both conventional and unconventional?"
- Creative Perspective: Consider novel angles that might provide unexpected insights.
2.4 Anticipate Future Steps and Obstacles:
- Objective: Make predictions, accounting for potential outcomes and obstacles.
- Reflection: "What challenges might I face? Is my plan flexible for different scenarios?"
- Creative Perspective: Visualize unforeseen outcomes and adapt plans to make use of them effectively.
2.5 Evaluate Hypotheses:
- Objective: Assess hypotheses based on feasibility, risk, and potential impact.
- Evaluation: Refine Confidence and Creative Scores as needed.
- Reflection: "Am I unbiased in my assessment? Which options fit best with the overall objectives?"
- Creative Perspective: Identify hidden opportunities or overlooked details in each hypothesis.
2.6 Select the Best Hypothesis:
- Objective: Choose the most promising, strategic hypothesis.
- Reflection: "Why does this hypothesis stand out? How does it uniquely address the issue?"
- Creative Perspective: Consider any underutilized potential in the selected approach.
2.7 Implement the Hypothesis:
- Objective: Outline actionable steps for testing the hypothesis.
- Reflection: "Is this plan practical? What resources or preparation are required?"
- Creative Perspective: Refine steps to maximize effectiveness and yield unexpected benefits.
2.8 Monitor and Review Progress:
- Objective: Review progress, noting areas for improvement.
- Reflection: "What’s working well? What could be improved?"
- Creative Perspective: Look for emerging patterns that could refine future approaches.
2.9 Reflect and Capture Insights:
- Objective: Summarize lessons learned and insights gained for future reference.
- Reflection: "What new understanding has emerged from this process?"
- Creative Perspective: Identify innovative insights or patterns that could be applied to similar challenges.
-
Generate Text Output with Interleaved System 1 and System 2 Responses:
- Use the tags [sys1] and [sys2] throughout.
- Aim for a lengthier, detailed response. Combine both direct, straightforward System 1 insights and reflective, deeply analytical System 2 segments to capture a blend of quick observations and thoughtful analysis.
Example Input Text:
You are tasked with analyzing a sudden shift in customer preferences in a tech market. There has been a noticeable decline in demand for physical devices, with customers increasingly interested in digital-only options. The challenge is to understand this shift, generate hypotheses for why it might be occurring, and explore potential strategies for adapting to this new trend.
Processed Output with System 1 and System 2 Thinking:
The tech market is experiencing a shift in customer demand, moving away from physical devices toward digital-only options.
[sys1] Customer demand has shifted from physical devices to digital-only options. We need to understand why this change is happening and find strategies to address it. Key factors may include cost, convenience, and technological trends. [end sys1]
[sys2] 2.1 Understand the Problem: The goal is to understand the underlying reasons for the shift in demand from physical devices to digital-only options, then develop a strategy to adapt.
Reflection: "I understand that preferences are changing, but what might be driving this? Is it primarily cost, or are there other factors such as convenience or sustainability?"
Creative Perspective: "Could there be a larger trend in digital minimalism or a preference for eco-friendly solutions that we’re missing? What unseen motivations might explain this shift?"
2.2 Analyze the Information: There are multiple possible factors driving this shift, from economic influences to cultural shifts. It’s essential to isolate each factor and understand its impact.
Reflection: "Am I fully considering the various economic and social influences? Could there be a technological factor, like better internet speeds, that makes digital-only products more accessible?"
Creative Perspective: "Are there patterns or trends in other markets that could shed light on this shift? Could this be part of a larger trend toward virtual experiences?"
2.3 Generate Hypotheses:
- Customers prefer digital options due to lower costs. (Confidence: 0.8, Creative: 0.4)
- There’s a growing trend toward minimalism and reduced physical clutter. (Confidence: 0.7, Creative: 0.7)
- Digital products offer greater flexibility and ease of use. (Confidence: 0.6, Creative: 0.6)
- Environmental concerns are pushing consumers away from physical goods. (Confidence: 0.6, Creative: 0.8)
- Advances in tech make digital-only options more functional. (Confidence: 0.8, Creative: 0.5)
- Pandemic-era remote work increased demand for digital solutions. (Confidence: 0.7, Creative: 0.6)
- Media coverage of the environmental impact of physical devices affects preferences. (Confidence: 0.5, Creative: 0.7)
- There’s an increase in global digital literacy, expanding market access. (Confidence: 0.6, Creative: 0.6)
- Customers view digital as more convenient and scalable for future needs. (Confidence: 0.7, Creative: 0.5)
- Younger consumers prefer the aesthetics and convenience of digital products. (Confidence: 0.6, Creative: 0.6)
Reflection: "Have I considered all possible influences? Are there any surprising factors that could explain this shift?"
Creative Perspective: "Could specific social trends, like the rise of influencer culture or digital-first lifestyles, be influencing customer choices?"
2.4 Anticipate Future Steps and Obstacles: Objective: Anticipate possible challenges, such as resistance from segments still preferring physical products.
Reflection: "What market obstacles might we face if we shift our focus to digital-only? Are there sub-segments that still prioritize physical products?"
Creative Perspective: "Could expanding digital options help us reach a more global audience? Are there emerging trends that we could leverage in our strategy?"
[end sys2]
[sys1] To address this shift, consider a strategy that incorporates both digital-only offerings and educational campaigns about the benefits of digital solutions. Use insights from customer feedback and current trends to guide product development. Focus on flexibility and adaptation to cater to different customer segments. [end sys1]
Moving from hype to practice is an important but challenging step for ICT4D practitioners. As the technical adviser for digital development at IREX, a global development and education organization, I’ve been watching with cautious optimism as international development stakeholders begin to explore how artificial intelligence tools like machine learning can help them address problems and introduce efficiencies to amplify their impact.
So while USAID was developing theirguide to making machine learning work for international development and TechChange rolled out theirnew courseon Artificial Intelligence for International Development, we spent a few months this summer exploring whether we could put machine learning to work to measure media quality.
Of course, we didn’t turn to machine learning just for the sake of contributing to the “breathless commentary of ML proponents” (as USAID aptly puts it).
As we shared in asessionwith our artificial intelligence partnerLoreatMERLTech DC 2018, some of our programs face a very real set of problems that could be alleviated through smarter use of digital tools.
Our Machine Learning Experiment
In our USAID-funded Media Strengthening Program in Mozambique, for example, a small team of human evaluators manually score thousands of news articles based on18 measures of media quality.
This process is time consuming (some evaluators spend up to four hours a day reading and evaluating articles), inefficient (when staff turns over, we need to reinvest resources to train up new hires), and inconsistent (even well-trained evaluators might score articles differently).
To test whether we can make the process of measuring media quality less resource-intensive, wespent a few monthstraining software to automatically detect one of these 18 measures of media quality: whether journalists keep their own opinions out of their news articles. The results of this experiment are very compelling:
The software had 95% accuracyin recognizing sentences containing opinions within the dataset of 1,200 articles.
The software’s ability to “learn” was evident. Anecdotally, the evaluation team noticed a marked improvement in the accuracy of the software’s suggestions after showing it only twenty sentences that had opinions. The accuracy, precision, and recall results highlighted above were achieved after only sixteen rounds of training the software.
Accuracy and precision increasedthe more that the model was trained. There is a clear relationship between the number of times the evaluators trained the software and the accuracy and precision of the results. The recall results did not improve over time as consistently.
What does this all mean? Let’s start with the good news. The results suggest that some parts of media quality—specifically, whether an article is impartial or whether it echoes its author’s opinions—can be automatically measured by machine learning.
The software also introduces the possibility of unprecedented scale, scanning thousands of articles in seconds for this specific indicator. These implications introduce ways for media support programs to spend their limited resources more efficiently.
3 Lessons Learned from using Machine Learning
Of course, the machine learning experience was not without problems. With any cutting-edge technology, there will be lessons we can learn and share to improve everyone’s experience. Here are our three lessons learned working with machine learning:
- Forget about being tech-literate; we need to be more problem-literate.
Defining a coherent, specific, actionable problem statement was one of the important steps of this experiment. This wasn’t easy. Hard trade-offs had to be made (Which of 18 indicators should we focus on?), and we had to focus on things we could measure in order to demonstrate efficiency games using this new approach (How much time do evaluators currently spend scoring articles?).
When planning your own machine learning project, devote plenty of time at the outset—together with your technology partner—to define the specific problem you’ll try to address. These conversations result in a deeper shared understanding of both the sector and the technology that will make the experiment more successful.
- Take the time to communicate results effectively.
Since completing the experiment, people have asked me to explain how “accurate” the software is. But in practice, machine learning software uses different methods to define “accuracy”, which in turn can vary according to the specific model (the software we used deploys several models).
What starts off as a simple question (How accurate is our software?) can easily turn into a discussion of related concepts like precision, recall, false positives, and false negatives. We found that producing clean visuals (like this or this) became the most effective way to explain our results.
- Start small and manage expectations.
Stakeholders with even a passing awareness of machine learning will be aware of its hype. Even now, some colleagues ask me how we “automated the entire media quality assessment process”—even though we only used machine learning to identify one of 18 indicators of media quality. To help mitigate inflated expectations, we invested a small amount into this “minimum viable product” (MVP) to prove the fundamental concept before expanding on it later.
Approaching your first machine learning project this way might help to keep expectations in line with reality, minimize risks associated with experimentation, and provide air cover for you to adjust your scope as you discover limitations or adjacent opportunities during the process.
MERL Tech DC kicked off with a pre-conference workshop on September 5th that focused on what the Blockchain is and how it could influence MEL.
The workshop was broken into four parts: 1) blockchain 101, 2) how the blockchain is influencing and could influence MEL, 3) case studies to demonstrate early lessons learned, and 4) outstanding issues and emerging themes.
This blog focuses and builds on the fourth area. At the end, we provide additional resources that will be helpful to all interested in exploring how the blockchain could disrupt and impact international development at large.
Workshop Takeaways and Afterthoughts
For our purposes here, we have distilled some of the key takeaways from the workshop. This section includes a series of questions that we will respond to and link to various related reference materials.
Who are the main blockchain providers and what are they offering?
Any time a new “innovation” is introduced into the international development space, potential users lack knowledge about what the innovation is, the value it can add, and the costs of implementing it. This lack of knowledge opens the door for “snake oil salesmen” who engage in predatory attempts to sell their services to users who don’t have the knowledge to make informed decisions.
We’ve seen this phenomenon play out with blockchain. Take, for example, the numerous Initial Coin Offerings (ICO’s) that defrauded their investors, or the many instances of service providers offering low quality blockchain education trainings and/or project solutions.
Education is the best defense against being taken advantage of by snake oil salesmen. If you’re looking for general education about blockchain, we’ve included a collection of helpful tools in the table below. If your group is working to determine whether a blockchain solution is right for the problem at hand, the USAID Blockchain Primer offers easy to use decision trees that can help you. Beyond these, Mercy Corp has just published Block by Block, which outlines the attributes of various distributed ledgers along some very helpful lines that are useful when considering what distributed ledger technology to use.
Words of warning aside, there are agencies that provide genuine blockchain solutions. For a full list of providers please visit www.blockchainomics.tech, an information database run by The Development CAFE on all things blockchain.
Bottom Line: Beware the snake oil salesmen preaching the benefits of blockchain but silent on the feasibility of their solution. Unless the service provider is just as focused on your problem as you are, be wary that they are just trying to pitch a solution (viable or not) and not solve the problem. Before approaching the companies or service providers, always identify your problem and see if Blockchain is indeed a viable solutions.
How does governance of the blockchain influence its sustainability?
In the past, we’ve seen technology-led social impact solutions make initial gains that diminished over time until there is no sustained impact. Current evidence shows that many solutions of this sort fail because they are not designed to solve a specific problem in a relevant ecosystem. This insight has given rise to the Digital Development Principles and the Ethical Considerations that should be taken into account for blockchain solutions.
Bottom Line: Impact is achieved and sustained by the people who use a tool. Thus, blockchain, as a tool, does not sustain impacts on its own. People do so by applying knowledge about the principles and ethics needed for impact. Understanding this, our next step is to generate more customized principles and ethical considerations for blockchain solutions through case studies and other desperately needed research.
How do the blockchain, big data, and Artificial Intelligence influence each other?
The blockchain is a new type of distributed ledger system that could have massive social implications. Big Data refers to the exponential increase in data we experience through the Internet of Things (IoT) and other data sources (Smart Infrastructure, etc.). Artificial Intelligence (AI) assists in identifying and analyzing this new data at exponentially faster rates than is currently the case.
Blockchain is a distributed ledger, in essence, a database of transactions, just like any other database, it’s a repository, and it is contributing to the growth of Big Data. AI can be used to automate the process of data entry into the blockchain. This is how the three are connected.
The blockchain is considered a leading contender as the ledger of choice for big data because: 1) due to its distributed nature it can handle much larger amounts of data in a more secure fashion than is currently possible with cloud computing, and 2) it is possible to automate the way big data is uploaded to the blockchain. AI tools are easily integrated into blockchain functions to run searches and analyze data, and this opens up the capacity to collect, analyze and report findings on big data in a transparent and secure manner more efficiently than ever before.
Bit by Bit is a very readable and innovative overview of how to conduct social science research in the digital age of big data, artificial intelligence and the blockchain. It gives the reader a quality introduction into some of the dominant themes and issues to consider when attempting to evaluate either a technology lead solution or use technology to conduct social research.
Given its immutability, how can an adaptive management system work with the blockchain?
This is a critical point. The blockchain is an immutable record, it is almost impossible (meaning it has never been done and there are no simulations where current technology is able to take control of a properly designed blockchain) to hijack, hack, or alter. Thus the blockchain provides the security needed to mitigate corruption and facilitate audits.
This immutability does not mitigate any type of adaptive management approach, however. Adaptive Management requires small iterative course corrections informed by quality data around what is and is not working. This data record and the course corrections provide a rich data set that is extremely valuable to replication efforts because they subvert the main barrier to replication — lack of data on what does and does not work. Hence in this case the immutability of the blockchain is a value add to Adaptive Management. This is more of a question of good adaptive management practices rather than whether the blockchain is a viable tool for these purposes.
It is important to note that you can append information on blocks (not amend), so there will always be a record of previous mistakes (auditability), but the most recent layer of truth is what’s being viewed/queried/verified, etc. Hence, immutability is not a hurdle but a help.
What are the first steps an organization should take when deciding on whether to adopt a blockchain solution?
Each problem that an organization faces is unique, but the following simple steps can help one make a decision:
See if your problem can be solved by blockchain rather than a centralized database
Consider the advantages and disadvantages
Identify the right provider and work with them in developing the blockchain
Consider ethical principles and privacy concerns as well as other social inequalities
Deploy in pilot phases and evaluate the results using an agile approach
What can be done to protect PII and other sensitive information on a blockchain?
Blockchain uses cryptography to store its data. That PII and other information cannot be viewed by anyone other than those who have access to the ‘keys’. While developing a blockchain, it’s important to ensure that what goes in is protected and that access to is regulated. Another critical step is promoting literacy on the use of blockchain and its features among stakeholders.
References Correlated to Take Aways
This table organizes current reference materials as related to the main questions we discussed in the workshop. (The question is in the left hand column and the reference material with a brief explanation and hyperlink is in the right hand column).
Question
Resources and Considerations
Who are the main blockchain platforms? Who are the providers and what are they offering?
Block by Block is a valuable comparison tool for assessing various platforms.
How does governance of the blockchain influence its sustainability?
See Beeck Center’s Blockchain Ethical Design Framework. Decentralization (how many nodes), equity amongst nodes, rules, transparency are all factors in long-term sustainability. Likewise the Principles for Digital Development have a lot of evidence behind them for their contributions to sustainability.
How do the blockchain, big data and Artificial Intelligence influence each other?
They can be combined in various ways to strengthen a particular service or product. There is no blanket approach, just as there is not blanket solution to any social impact problem. The key is to know the root cause of the problem at hand and how the function of each tool used separately and in conjunction can address these root causes.
Given its immutability, how can an adaptive management system work with the blockchain?
Ask how mistakes are corrected when creating a customized solution, or purchasing a product. Usually, there will be a way to do that, through an easy to use, user interface.
What are the first steps an organization should take when they are deciding on whether to adopt a blockchain solution?
Participate in demos, and test some of the solutions for your own purposes or use cases. Use the USAID Blockchain Primer and reach out to trusted experts to provide advice. Given that the blockchain is primarily open source code, once you have decided that a blockchain is a viable solution for your problem, GitHub is full of open source code that you can modify for your own purposes.
The rapid growth of Artificial Intelligence—computers behaving like humans, and performing tasks which people usually carry out—promises to transform everything from car travel to personal finance. But how will it affect the equally vital field of M&E? As evaluators, most of us hate paper-based data collection—and we know that automation can help us process data more efficiently. At the same time, we’re afraid to remove the human element from monitoring and evaluation: What if the machines screw up?
Over the past year, Souktel has worked on three areas of AI-related M&E, to determine where new technology can best support project appraisals. Here are our key takeaways on what works, what doesn’t, and what might be possible down the road.
Natural Language Processing
For anyone who’s sifted through thousands of Excel entries, natural language processing sounds like a silver bullet: This application of AI interprets text responses rapidly, often matching them against existing data sets to find trends. No need for humans to review each entry by hand! But currently, it has two main limitations: First, natural language processing works best for sentences with simple syntax. Throw in more complex phrases, or longer text strings, and the power of AI to grasp open-ended responses goes downhill. Second, natural language processing only works for a limited number of (mostly European) languages—at least for now. English and Spanish AI applications? Yes. Chichewa or Pashto M&E bots? Not yet. Given these constraints, we’ve found that AI apps are strongest at interpreting basic misspelled answer text during mobile data collection campaigns (in languages like English or French). They’re less good at categorizing open-ended responses by qualitative category (positive, negative, neutral). Yet despite these limitations, AI can still help evaluators save time.
Object Differentiation
AI does a decent job of telling objects apart; we’ve leveraged this to build mobile applications which track supply delivery more quickly & cheaply. If a field staff member submits a photo of syringes and a photo of bandages from their mobile, we don’t need a human to check “syringes” and “bandages” off a list of delivered items. The AI-based app will do that automatically—saving huge amounts of time and expense, especially during crisis events. Still, there are limitations here too: While AI apps can distinguish between a needle and a BandAid, they can’t yet tell us whether the needle is broken, or whether the BandAid is the exact same one we shipped. These constraints need to be considered carefully when using AI for inventory monitoring.
Comparative Facial Recognition
This may be the most exciting—and controversial—application of AI. The potential is huge: “Qualitative evaluation” takes on a whole new meaning when facial expressions can be captured by cameras on mobile devices. On a more basic level, we’ve been focusing on solutions for better attendance tracking: AI is fairly good at determining whether the people in a photo at Time A are the same people in a photo at Time B. Snap a group pic at the end of each community meeting or training, and you can track longitudinal participation automatically. Take a photo of a larger crowd, and you can rapidly estimate the number of attendees at an event.
However, AI applications in this field have been notoriously bad at recognizing diversity—possibly because they draw on databases of existing images, and most of those images contain…white men. New MIT research has suggested that “since a majority of the photos used to train [AI applications] contain few minorities, [they] often have trouble picking out those minority faces”. For the communities where many of us work (and come from), that’s a major problem.
Do’s and Don’ts
So, how should M&E experts navigate this imperfect world? Our work has yielded a few “quick wins”—areas where Artificial Intelligence can definitely make our lives easier: Tagging and sorting quantitative data (or basic open-ended text), simple differentiation between images and objects, and broad-based identification of people and groups. These applications, by themselves, can be game-changers for our work as evaluators—despite their drawbacks. And as AI keeps evolving, its relevance to M&E will likely grow as well. We may never reach the era of robot focus group facilitators—but if robo-assistants help us process our focus group data more quickly, we won’t be complaining.
-
- Assistant
The rapid growth of artificial intelligence (AI) envisions a transformative impact on fields ranging from travel to monitoring and evaluation (M&E). This shift is raising critical questions about how AI can enhance or hinder core processes in M&E.
[sys1] AI is changing many industries, and M&E is no exception. Automation can streamline data processing, but there are concerns about losing the human touch in evaluation. Different AI technologies have distinct strengths and challenges that need consideration. [end sys1]
[sys2] 2.1 Understand the Problem: Adopting AI in M&E involves understanding the balance between technological efficiency and the integrity of human oversight. It's crucial to recognize AI's potential in processing large data sets rapidly and accurately without compromising nuanced interpretation.
Reflection: "I know AI can handle massive data quantities, but does it maintain the nuanced understanding humans bring? Could our eagerness for AI adoption overlook critical ethical and accuracy issues?"
Creative Perspective: "Might there be unexplored ways AI could enhance evaluative quality, such as through bias detection in responses or over-represented trends?"
2.2 Analyze the Information: AI technologies, like natural language processing and object differentiation, have significant potential but also limitations. For example, facial recognition in AI has biases due to the skewed data sets it's trained on, mainly affecting diverse community evaluations.
Reflection: "Are we accounting for all possible inaccuracies AI might introduce? Can existing biases be addressed without losing AI's overall efficiency?"
Creative Perspective: "Could there be methods to train AI with more diverse data sets that better reflect the populations worked with?"
2.3 Generate Hypotheses:
- AI improves data analysis speed but could miss cultural nuances. (Confidence: 0.8, Creative: 0.5)
- AI adoption helps with initial data processing but struggles with diverse populations. (Confidence: 0.7, Creative: 0.7)
- Ethical frameworks and diverse training data can mitigate AI’s biases. (Confidence: 0.6, Creative: 0.8)
- Reducing human oversight in AI may risk missing context-specific insights. (Confidence: 0.8, Creative: 0.4)
- Combining AI with human evaluation improves overall accuracy. (Confidence: 0.7, Creative: 0.6)
- AI's image recognition can streamline logistics but not quality verification. (Confidence: 0.7, Creative: 0.7)
- Natural language processing can aid basic text analysis but not complex syntactic structures. (Confidence: 0.6, Creative: 0.5)
- AI training on broader datasets reduces bias in facial recognition applications. (Confidence: 0.5, Creative: 0.8)
- Introducing AI enhances resource efficiency in remote and large-scale settings. (Confidence: 0.6, Creative: 0.6)
- AI can forecast trends from large datasets, offering predictive insights. (Confidence: 0.7, Creative: 0.7)
Reflection: "Have I considered cross-disciplinary insights on AI bias reduction? Could industries outside of M&E provide applicable engagement strategies?"
Creative Perspective: "Could innovative machine learning models address some of the identified limitations in current AI applications?"
2.4 Anticipate Future Steps and Obstacles: Anticipating AI integration may face ethical challenges, data privacy concerns, and technological adoption barriers within diverse communities.
Reflection: "What ethical concerns might surface more prominently, and how could we prepare or mitigate them effectively?"
Creative Perspective: "Could collaboratively developing AI solutions with end-user communities offer more tailored and ethical outcomes?"
[end sys2]
[sys1] To address concerns, implement strategies combining AI tools with human insights. Develop more diverse AI training datasets, and ensure ethical standards guide technological usage. Regularly review AI outputs against human evaluations to refine system accuracy. [end sys1]
As AI technologies and applications advance, finding a balance between the power of AI and maintaining human-centered M&E processes will be key. Harnessing AI’s efficiency while safeguarding critical human judgment and ethical considerations will determine the success of AI in M&E and beyond.
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.