KeithB18

Vern, Do you think the DPA could be used to force a Quaker owned textile business to supply uniforms to the DOD? Do you think it is possible that invoking DPA might raise some compelling questions about the nature of freedom, including freedom of speech and worship, and ownership of one's labor?

I think Anthropic has every right to not do business with DOD, if that's what they want to do. However, I don't think you can sell your AI to the DOD and not expect it to be used in ways you never intended. How far is it, really, from using AI to improve the solicitation that buys bombs to using AI to drop those bombs? I don't know. Like you said, Vern, where's the line? Can DPA force them to sell? I am no expert in DPA, but if a company has behaved in such a way that it is believable that they are a corporate conscientious objector, I'd have a hard time forcing them to do so. But Anthropic has not behaved that way, so I think the Government would probably prevail. Should Americans take that position? Many do. When I was active duty from 02-06, I worked with a lot of DOD civilians who were against the Iraq war. Ultimately, it is an individual decision. After my AD experience in Iraq and Afghanistan and a few years as a DOD contractor, I decided I didn't want to have anything to do with the DOD,* so I worked in civilian agencies, and still do. It could be that I'm as hypocritical as Anthropic appears to be--it's all the same government right? *It wasn't just the activities DOD involves itself in, it was also the stifling bureaucracy that pushed me out of DOD.

I've seen this have some really perverse effects. One organization I worked for--doing 90% commercial services buys--had GS-15s as 25% of its workforce. Just because you can hire at a more senior grade, doesn't mean you should. That said, no government manager has ever gotten credit for bringing the wage bill down, so there's little incentive to try to do so.

I'd go so far to say that the pursuit of full and open competition is actively harmful in a significant number of procurements. It might even be a majority. For example, I'm working on basic and applied R&D right now and there are basically no cases where prices go down and quality goes up because of competition. This is something I would like to think about and think about how to test.

Here Here's the picture, anyhow.

What I'm trying to say is that there is a false expectation that the output of our evaluation and selection process is an objectively correct answer. We can change our processes in ways that make it more likely we've selected a correct answer (there could be many), and I've exchanged posts with you on ways I think we should do that--capabilities based assessments; oral presentations preferred to written responses, much of which I learned from reading your writings. But it won't get close unless we continually interrogate it, and also I don't think anyone here is suggesting otherwise. Maybe this isn't much of a radical position. I would say, however, I have recent experience that suggests some decision makers rely on the process as a way to avoid criticism and discretion. To tie it back to the original subject of this thread--relying on process without being intensely critical of the process itself is similar relying on AI. As an aside: Attaching a picture I took moments ago of decision books recommended by Vern. Can't say I've read either cover to cover, but I have referred to them over the years. To some extent we are all Camus' Sisyphus perpetually rolling the boulder up a hill.

When you bought your house (most people's largest purchase) did you: Set evaluation criteria that could not change? Ask for a written proposal from each house for sale without, potentially, even stepping inside the house yourself? Have 3, 5, 7, or more extended family members review each house against the evaluation criteria and make a recommendation to you, as the selection authority? Document the reasons you made your decisions and the tradeoffs you made? (As an aside, when my wife and I started looking for our home, we thought we wanted three bedrooms and a basement. We bought a home with two bedrooms and no basement. We chose this house because it was very close to metro and a grocery store. Our initial evaluation criteria were incorrect! Happens all the time--people are not always very good at knowing what is important to them.) What about when you picked your spouse? Did you write down the tradeoffs you made? (I hope not!) Do you shop around to get a better deal on your dry-cleaning or just go to the most convenient one? I would posit that there are very few commercial, daily transactions where we make more than the most basic tradeoffs. Often they are only implicit: "I don't want to pay that price for that can of green beans." So while it is superficially true that tradeoffs are made at the individual and government levels, the form those tradeoffs take is so significantly different that it constitutes a completely different thing.

The definition of "qualitative" via google is "relating to, measuring, or measured by the quality of something rather than its quantity." You can assign a number to quality, but I think the assigned number is, at least in part, an expression of the assigner's values.

This answer isn't going to be satisfying, because it is not satisfying to me. It's something I'm still working on. I've been influenced lately by Michael Polanyi who wrote a lot about implicit and tacit knowledge. He wrote, "You know more than you can say." His works are kind of difficult, at least for me. I'm working in a basic and applied R&D context right now, and that line resonates with me when you have very experienced and top of their field scientists reviewing proposals. They may, quite rightly imo, rely on their intuition in selecting projects. Intuition doesn't necessarily lend itself to writing strengths and weaknesses. Qualitative decisions are non-numeric judgments, and a lot of judgments are based on values. Values often operate in the background, implicitly. (There's a study on "Terror Management Theory" that showed that judges rule more harshly against people that violate the judges values when they are reminded of their own mortality beforehand: https://psycnet.apa.org/doiLanding?doi=10.1037%2F0022-3514.57.4.681. Which is just to point out that values can operate in unseen and strange ways.) I'm not sure how to translate this into a normative rule. What I'm trying to do is what Oliver Wendell Holmes was doing in "The Path of the Law." Get the dragon out into the daylight, so that "you can count his teeth and claws, and see just what is his strength." I'm trying to understand what's really going on when procurement decisions are made, and adjust accordingly. Like, is there any other decision making context, in the entirety of human experience, that works like Federal procurement decisions do?

I have done some experiments with ChatGPT on SOWs and solicitations. It does okay with SOWs--they still need a fair bit of tailoring but I'd say it can produce a decent starting point. It cannot produce a reasonable solicitation yet, but there may be some series of prompts you could use to get a better output for the solicitation (I haven't explored it enough). So I think it will eventually be useful for those documents, but my rule of thumb is the more I relied on the AI tool, the more editing and oversight the document needs. It may not result in a reduction in effort, rather a shift in what you spend your effort on. I also wonder if this shift is even worth it...I guess we'll see. That said, I have heard colleagues talk about using AI for evaluation, and that's where I get concerned. A government of the people needs people making decisions on behalf of the people. An idea I've been working on over the past few weeks is that we use contracting processes (and other government processes) to obscure the discretion given to deciders, mostly (probably? I'm still working through it) because deciders are afraid of criticism or don't want to decide.* To delegate decision making to AI is a further dereliction of duty. *This works when you need the lowest priced #2 pencil, but treating the evaluation of a complex professional service as if it is a math problem to be solved isn't how qualitative decisions are made.

To answer the question: Yes. We have all these fancy tools and filling them in daily/weekly/monthly becomes a significant part of the job. Then no one ever looks at the tools themselves, so you get one-off data calls that are already answered by looking in the tool. Then a persistent comment at your weekly or monthly staff meeting is to update the information in the tool. All of this to give management a feeling that they are managing something. Back in my SES days, I really tried hard to find the information myself before asking a staff member for it. The amount of work (and sometimes panic) that comes with a data call is massive. There is so much information already at your finger tips that you rarely need to ask if you try to know where to look. Even something basic like usaspending.gov will give you basic information on the project, vendor, obligation, POP, and outlays. (As an aside, I read a long article about Skunk Works yesterday. They were able to field an SR-71 prototype in fewer than 200 days. Amazing.)

It's definitely weird and the whole thing was entirely avoidable with better management (It's no one's fault, everyone involved was doing their best with the facts as they understood them at the time.) I have come into the problem very late--two plus years late. It's 3 year research and development money. Generally, R&D is non-severable, but these funds were obligated on a consortium management other transaction. I'm squinting hard at it but I'm struggling to conclude that services related to managing a consortium are non-severable. The intent of the funding was to fund R&D projects through the consortium, but the projects didn't materialize until basically right now. So, a three year appropriation made in FY22. Funds obligated in FY23. It's now FY26. I think the obligation was properly made in the period of availability and that the consortium management firm can invoice against the funds. Others disagree.

I'm definitely interpreting the Redbook differently (potentially incorrectly) than others. (I've scoured it over the last couple of days.) For example, page 5-7 of the Redbook says: "Apart from the extended period of availability, multiple year appropriations are subject to the same principles applicable to annual appropriations and do not present any special problems." A principle of annual appropriations is that there is a one year exception that allows severable services to cross fiscal years. I would argue that the one year exception is tied to the period of availability of the appropriation, not the cardinal number "1." If you follow this logic, an annual appropriation attaches a one year exception while a three year appropriation attaches a three year exception. At this point the attorneys point me to 41 USC 3902. It says what it says. But I having trouble understanding why Congress would appropriate funds for three years, offering the agency significant flexibility, and then restrict that flexible appropriation in this manner. Essentially, in the last year of its period of availability, three year money becomes annual money (when contracting for severable services). Maybe that really is the intent, but I don't think it makes sense. (A lot of things don't make sense though.) I run into the same problem reading B-317636, which concludes: "...federal agencies may use multiple year or no-year funds to enter into contracts for severable services for a period of performance longer than 1 year and an agency using multiple year or no-year funds is free to contract for the full period of availability the statute appropriating those funds allows." Is the "full period of availability of the statute appropriating those funds" a cardinal number or is it the specific fiscal year(s) called for in the law? If it is the latter, a lot of us have been sloppy with our language in saying things are "3 year money." They aren't, they are FY2024, 2025, and 2026 monies (for example). I definitely agree that bona fide need attaches to OTs. But I also think the analysis that GAO did in the case referenced by C_Culham is misguided. They applied grant and procurement principles in order to analyze OTs. They said "This one looks more like a grant and that one looks more like a contract and we'll evaluate them on those grounds." That kind of thinking is just going to make OTs grants or procurement contracts. OTs need to be thought of as separate thing outside of grants and procurement. (Recognizing there is a lot of debate about OTs and their use, but if they are going to be a thing, they need to be a different thing than what we already have.)

"Most want to know, but too many don't like to learn." I'm going to be thinking about that for awhile. How many of us are unwilling to adjust our priors in the face of new evidence? I certainly try to do that but I have an ego and pride too. It's not always easy to learn, but you have to commit to it, I think.

I have to preface this with the fact that agency lawyers have told me I am wrong about my interpretation of this issue. But if I've learned anything after being around this forum for 15 or so years, I don't always trust agency lawyer interpretation. That said, I'm 85% sure I'm wrong about this and the lawyers are right. But I want to pose the question here to double check. Question One: You have funds appropriated for three years (3 year money). Can you issue a contract for severable services on the last day of the third year of availability with a period of performance for three years? Attorneys point me to 41 USC 3902, which, to be frank, limits the period of performance for severable services that cross fiscal years to one year. But I don't think that statute makes a lot of sense where you have multiple year appropriations. Did Congress intend to limit 3 year appropriations in that manner? My reading of B-317636 makes me wonder. If annual appropriations have a one year severable services limit, why wouldn't three year appropriations come with a three year limit? (Again, willing to be wrong here, but I'm struggling with the logic.) Question Two: Same scenario, but it is an other transaction instead of a procurement contract. Does that change the analysis? It seems clear to me that 41 USC 3902 wouldn't apply here, but bona fide need would. But what's the bare minimum an agency needs to do to comply with bona fide need? I am not even willing to concede that severability need to be read into other transactions.