You walk into a hearing expecting a neutral judge who will listen to both sides. Instead, you find a judge who spent months building your case file—ordering exams, gathering records, forming preliminary views. Is this a fair hearing or a predetermined outcome?
This episode explores the Social Security disability system, the largest adjudication system in the United States, where administrative law judges both develop the evidence and decide the case. We contrast adversarial and inquisitorial models of justice, examine why the U.S. selectively borrowed from civil law systems without their safeguards, and unpack the cognitive risks—confirmation bias, ownership effects, and implicit prejudgment—that emerge when investigation and adjudication combine in a single person.
We also grapple with a harder question: Was this system designed wrong, or is it actually the more humane choice for vulnerable claimants who can't afford lawyers and wouldn't survive a fully adversarial fight?
Key Concepts
- Adversarial vs. Inquisitorial Systems: The fundamental difference between party-driven investigation (U.S. courts) and judge-driven investigation (civil law systems), and why mixing them creates new problems
- The Duty to Develop the Record: How ALJs are required to gather evidence to ensure claims are fairly decided, even when claimants are unrepresented
- Confirmation Bias in Adjudication: Why forming preliminary views while building the record creates a cognitive loop that's difficult to escape
- Structural Bias vs. Individual Bias: The difference between proving an individual judge is biased and identifying systemic risks in how roles are combined
- Consultative Examinations: How ALJ-ordered medical exams can both help and hurt claimants, depending on what they reveal
Cases & Regulations Discussed
Withrow v. Larkin, 421 U.S. 35 (1975)
- Establishes that combining investigative and adjudicative functions doesn't automatically violate due process
- Creates a high bar: challengers must prove "actual bias" or overcome the "presumption of honesty and integrity"
- Sets the constitutional framework that allows Social Security's current structure
20 C.F.R. § 404.944 - ALJ's Duty to Develop the Record
- Codifies the affirmative obligation to develop a complete record
- Requires ALJs to request additional evidence, seek records, and order consultative exams when needed
Episode Highlights
Why This Matters: Over 500,000 disability hearings occur annually—more than the entire federal court system combined. Most claimants have already been denied twice and are unrepresented. The structure of these hearings determines whether they get a fair shot.
The Design Choice: The system wasn't broken by accident—it was intentionally designed to help claimants who couldn't navigate an adversarial process. The alternative would be government attorneys arguing against every disability applicant.
The Cognitive Problem: Even well-intentioned ALJs face predictable psychological risks when they develop evidence and then judge it. The bias isn't malicious—it's structural.
The Constitutional Standard: Courts have upheld this structure because proving "actual bias" is nearly impossible when the bias operates through investigation choices rather than overt prejudice.
00:00:14 --> 00:00:17 Hello, and welcome to Administrative Remedies,
00:00:17 --> 00:00:19 because you can't fix what you don't understand.
00:00:19 --> 00:00:21 Brought to you in part by the University of Tulsa
00:00:21 --> 00:00:24 College of Law. I'm Gwendolyn Savitz, an associate
00:00:24 --> 00:00:26 professor here at TU and the associate dean of
00:00:26 --> 00:00:28 research and intellectual life. And I'm Mark
00:00:28 --> 00:00:31 Rourke. I'm the dean of the College of Law. We'll
00:00:31 --> 00:00:33 be breaking down complex doctrines with real
00:00:33 --> 00:00:35 -life analogies and examples to demystify the
00:00:35 --> 00:00:37 world of administrative law for everyone trying
00:00:37 --> 00:00:39 to understand how government actually works.
00:00:40 --> 00:00:41 Agencies are the main way the federal government
00:00:41 --> 00:00:44 gets things done. It's not through Congress for
00:00:44 --> 00:00:46 reasons we'll be addressing over the course of
00:00:46 --> 00:00:50 this series. We are going to continue our discussion
00:00:50 --> 00:00:54 of adjudication. So, Mark, let's imagine you
00:00:54 --> 00:00:56 are in a legal dispute. You walk into a courtroom.
00:00:57 --> 00:00:59 In this courtroom, each side has its own lawyer.
00:00:59 --> 00:01:02 Your lawyer gathered evidence and built your
00:01:02 --> 00:01:04 case. The other side's lawyer did the same thing.
00:01:04 --> 00:01:07 You have a judge sitting at the bench. The judge
00:01:07 --> 00:01:09 didn't investigate anything. It's just going
00:01:09 --> 00:01:12 to listen to both sides and to suck. Yeah, that's
00:01:12 --> 00:01:14 how it's supposed to work. I mean, we teach that
00:01:14 --> 00:01:16 in law school. That's law school 101. That's
00:01:16 --> 00:01:20 how the court process works. OK, so let's imagine
00:01:20 --> 00:01:22 a different courtroom. So this time you walk
00:01:22 --> 00:01:24 in. There's no opposing counsel. There's just
00:01:24 --> 00:01:27 a judge. And that judge spent the last six months
00:01:27 --> 00:01:29 reviewing your records, ordering examinations,
00:01:29 --> 00:01:31 gathering evidence, informing opinions about
00:01:31 --> 00:01:35 your credibility. Well, I mean. That's not as
00:01:35 --> 00:01:37 unusual as you're making it sound. I mean, judges
00:01:37 --> 00:01:40 review extensive filings before trial all the
00:01:40 --> 00:01:42 time. They read depositions. They look at exhibits.
00:01:42 --> 00:01:45 They roll in pretrial motions. By the time the
00:01:45 --> 00:01:47 arguments start, a judge has already absorbed
00:01:47 --> 00:01:50 a ton of evidence. It's natural they might have
00:01:50 --> 00:01:53 opinions. OK, yeah, fair point. But there really
00:01:53 --> 00:01:55 is an important distinction. In a normal case,
00:01:55 --> 00:01:57 the judge is reading what the parties chose to
00:01:57 --> 00:01:59 put in front of them. Each side decided what
00:01:59 --> 00:02:02 to file, what to emphasize, what to leave out.
00:02:02 --> 00:02:03 The judge is absorbing the party's evidence.
00:02:04 --> 00:02:06 In the other courtroom, the judge is deciding
00:02:06 --> 00:02:09 what evidence to gather. Like, the judge might
00:02:09 --> 00:02:11 have ordered medical exams or requested specific
00:02:11 --> 00:02:14 records. It's the judge in this courtroom that
00:02:14 --> 00:02:16 shaped the universe of evidence before the hearing
00:02:16 --> 00:02:19 even started. And then that same judge sits down,
00:02:19 --> 00:02:21 questions you, and decides your case. But the
00:02:21 --> 00:02:24 difference isn't just that they've seen the evidence,
00:02:24 --> 00:02:27 it's that they built the evidence. Yes, they
00:02:27 --> 00:02:29 built it and then they judge it. That is social
00:02:29 --> 00:02:33 security disability. That's how the largest adjudication
00:02:33 --> 00:02:36 system in the U .S. actually works. How large
00:02:36 --> 00:02:39 are we talking? Over half a million hearings
00:02:39 --> 00:02:42 a year. That's more cases than the entire federal
00:02:42 --> 00:02:44 court system combined. And in every one of those
00:02:44 --> 00:02:47 hearings, the judge who decides the case is the
00:02:47 --> 00:02:48 same person who helped build the record the case
00:02:48 --> 00:02:51 is resting on. OK, that cannot be how it was
00:02:51 --> 00:02:54 designed to work, though. It is absolutely how
00:02:54 --> 00:02:57 it was designed to work. But before anyone panics,
00:02:57 --> 00:02:59 this wasn't an accident. It wasn't a power grab.
00:02:59 --> 00:03:02 It was a deliberate choice that was made for
00:03:02 --> 00:03:04 reasons that make sense when you understand the
00:03:04 --> 00:03:08 history. So why does this feel so wrong? Because
00:03:08 --> 00:03:10 it violates every instinct that people develop
00:03:10 --> 00:03:12 in law school about what a fair hearing should
00:03:12 --> 00:03:15 actually look like. And today we're going to
00:03:15 --> 00:03:18 sit with that discomfort and try and understand
00:03:18 --> 00:03:21 where it comes from and what the law has done
00:03:21 --> 00:03:23 to try and answer it. Before we go further, we
00:03:23 --> 00:03:25 should lay out some basics for the listeners.
00:03:26 --> 00:03:28 You know civil law. What are the two main models
00:03:28 --> 00:03:31 for how courtrooms work? Okay. In the United
00:03:31 --> 00:03:34 States and most common law countries. We use
00:03:34 --> 00:03:36 the adversarial system, two sides, each with
00:03:36 --> 00:03:39 their own lawyer. Each side investigates independently,
00:03:39 --> 00:03:41 gathers their own evidence. They interview their
00:03:41 --> 00:03:44 own witnesses. They build their own case. We
00:03:44 --> 00:03:47 put all of the costs on the litigants for hiring
00:03:47 --> 00:03:50 their own attorney. And so then they present
00:03:50 --> 00:03:52 this to a judge or jury who decides based on
00:03:52 --> 00:03:55 what they've heard. And the judge basically serves
00:03:55 --> 00:03:57 as a referee. They enforce the rules of evidence.
00:03:57 --> 00:04:00 They manage the proceedings. They rule on objections.
00:04:00 --> 00:04:02 But the judge doesn't really go out and gather
00:04:02 --> 00:04:05 evidence. The judge works with what the parties
00:04:05 --> 00:04:07 bring. Okay. So what's the other model? It's
00:04:07 --> 00:04:10 called the inquisitorial model. It's used across
00:04:10 --> 00:04:13 Europe. Most of Latin America, parts of Asia,
00:04:13 --> 00:04:16 very different philosophically. The judge is
00:04:16 --> 00:04:19 very active. The judge investigates. He gathers
00:04:19 --> 00:04:22 evidence, questions witnesses. He develops the
00:04:22 --> 00:04:24 record. So in this version, it's the judge running
00:04:24 --> 00:04:27 the investigation. Yeah. The idea is that the
00:04:27 --> 00:04:29 judge is searching for truth, not just refereeing
00:04:29 --> 00:04:33 a contest between two sides. And in some ways,
00:04:33 --> 00:04:35 it's arguably fair. The outcome doesn't depend
00:04:35 --> 00:04:37 as heavily on who can afford better lawyers,
00:04:37 --> 00:04:40 right? This is the whole premise of crime and
00:04:40 --> 00:04:42 punishment from Fedor Dostoevsky. The judge has
00:04:42 --> 00:04:44 resources to investigate that a poor party might
00:04:44 --> 00:04:47 not have. If you can't afford to hire an expert
00:04:47 --> 00:04:49 or track down a witness, the investigating judge
00:04:49 --> 00:04:52 can do that. OK, what safeguards are built into
00:04:52 --> 00:04:55 this system? Oh, significant ones, right? In
00:04:55 --> 00:04:58 many civil law systems, the investigating judge
00:04:58 --> 00:05:00 and the deciding judge are different people.
00:05:00 --> 00:05:03 There are multiple judges reviewing cases at
00:05:03 --> 00:05:05 each stage. There are strict procedural rules
00:05:05 --> 00:05:07 governing what the investigation looks like.
00:05:07 --> 00:05:10 And the investigation function is checked by
00:05:10 --> 00:05:12 institutional design. That's where the American
00:05:12 --> 00:05:15 borrowing gets interesting and problematic. We
00:05:15 --> 00:05:18 selectively imported some of these inquisitorial
00:05:18 --> 00:05:20 features for specific problems, but we didn't
00:05:20 --> 00:05:23 bring over the safeguards. What did we take?
00:05:23 --> 00:05:25 We took the active judge, a judge who's developing
00:05:25 --> 00:05:27 the record, who's ordering examinations, who's
00:05:27 --> 00:05:30 gathering evidence. Social Security disability
00:05:30 --> 00:05:31 is the clearest example. That's what we were
00:05:31 --> 00:05:34 talking about. But then we grafted that onto
00:05:34 --> 00:05:37 the American framework with a single ALJ who's
00:05:37 --> 00:05:39 supposed to be neutral, but who's also developing
00:05:39 --> 00:05:41 the evidence. There's no separate investigating
00:05:41 --> 00:05:43 judge. There's no panel review at the hearing
00:05:43 --> 00:05:46 stage. It's a hybrid, and hybrids can inherit
00:05:46 --> 00:05:49 the risk of both parents. Let's walk through
00:05:49 --> 00:05:51 a little more detail about what happens in a
00:05:51 --> 00:05:53 Social Security disability case. Since this is
00:05:53 --> 00:05:55 where most people are going to encounter this
00:05:55 --> 00:05:58 type of system. And because the gap between what
00:05:58 --> 00:06:00 they expect and what actually happens can be
00:06:00 --> 00:06:03 enormous. Okay. Lay it on me. All right. You
00:06:03 --> 00:06:06 apply for disability benefits. You're claiming
00:06:06 --> 00:06:08 you can't work. You've had a back injury or mental
00:06:08 --> 00:06:11 illness or chronic pain, whatever it is. Social
00:06:11 --> 00:06:13 Security denies your claim at the initial level.
00:06:14 --> 00:06:16 You appeal. It's denied again at reconsideration.
00:06:17 --> 00:06:19 We talked about this earlier. And then you request
00:06:19 --> 00:06:21 a hearing before an administrative law judge.
00:06:21 --> 00:06:24 OK. So far, that sounds like a normal appeals
00:06:24 --> 00:06:26 process. It is. And this is where it changes.
00:06:26 --> 00:06:29 Before that hearing ever happens, the ALJ gets
00:06:29 --> 00:06:32 your entire file, the medical records, the agency
00:06:32 --> 00:06:35 evaluations from before, the prior denials, the
00:06:35 --> 00:06:38 consultative exam reports, everything. We're
00:06:38 --> 00:06:40 not talking a summary. We're talking the whole
00:06:40 --> 00:06:43 thing, which can be hundreds or even thousands
00:06:43 --> 00:06:45 of pages. So the judge walks in having already
00:06:45 --> 00:06:48 read the whole case. More than just read it.
00:06:48 --> 00:06:50 The ALJ isn't just passively absorbing what's
00:06:50 --> 00:06:52 there. They can order additional medical exams.
00:06:53 --> 00:06:55 They can request more records. They can ask for
00:06:55 --> 00:06:57 opinions from medical or vocational experts.
00:06:57 --> 00:07:00 They're actively shaping what evidence exists.
00:07:00 --> 00:07:02 And then the hearing happens. Yeah. You show
00:07:02 --> 00:07:05 up. Maybe you have a lawyer. Maybe you don't.
00:07:05 --> 00:07:08 About half of claimants at the hearing are unrepresented.
00:07:08 --> 00:07:11 You testify. The ALJ asks you some questions,
00:07:11 --> 00:07:14 not... as a blank slate, but with views that
00:07:14 --> 00:07:16 are already formed about what the evidence says
00:07:16 --> 00:07:19 that they helped assemble. I want to push back
00:07:19 --> 00:07:21 on that for a second. Judges form preliminary
00:07:21 --> 00:07:23 views in every single case. An appellate judge
00:07:23 --> 00:07:26 reads the briefs, has tentative view before oral
00:07:26 --> 00:07:28 argument. We like to say sometimes that it seems
00:07:28 --> 00:07:30 like a judge makes decisions based on what's
00:07:30 --> 00:07:34 in a cereal bowl, right? But a trial judge reads
00:07:34 --> 00:07:36 pretrial motions. He knows what's coming. Is
00:07:36 --> 00:07:38 this really that different? We mentioned it a
00:07:38 --> 00:07:41 little, but it's really worth repeating. This
00:07:41 --> 00:07:44 is different in a really critical way. In a traditional
00:07:44 --> 00:07:46 case, the judge is forming their views based
00:07:46 --> 00:07:48 on evidence that the parties have chosen to present.
00:07:49 --> 00:07:52 Here, the Social Security ALJ is forming opinions
00:07:52 --> 00:07:55 based at least partly on the evidence that they
00:07:55 --> 00:07:57 chose to develop. That is not the same thing.
00:07:58 --> 00:07:59 When you're the one who decides what evidence
00:07:59 --> 00:08:02 to gather, your preliminary views aren't just
00:08:02 --> 00:08:04 shaped by the record. They are baked into it.
00:08:04 --> 00:08:07 Okay, that is a real distinction. Why exactly
00:08:07 --> 00:08:10 did we set this up this way? The system was intentionally
00:08:10 --> 00:08:13 built like this at a time when the primary concern
00:08:13 --> 00:08:16 was the opposite problem. When Social Security
00:08:16 --> 00:08:19 disability was taking shape in the 1950s and
00:08:19 --> 00:08:21 then expanding through the 60s and 70s, the worry
00:08:21 --> 00:08:23 wasn't that judges would investigate claimants
00:08:23 --> 00:08:26 too aggressively. It was that legitimate claims
00:08:26 --> 00:08:28 would be denied because the claimants, most of
00:08:28 --> 00:08:31 whom would be poor and unsophisticated and unrepresented,
00:08:32 --> 00:08:34 wouldn't be able to navigate the system well
00:08:34 --> 00:08:36 enough to prove their case. So the judge stepping
00:08:36 --> 00:08:38 in was meant to level the playing field. Exactly.
00:08:38 --> 00:08:41 The ALJ has an affirmative duty to make sure
00:08:41 --> 00:08:43 the record is complete. They're supposed to gather
00:08:43 --> 00:08:46 all the evidence needed to decide the case fairly,
00:08:46 --> 00:08:48 even if the claimant doesn't know to ask for
00:08:48 --> 00:08:52 that evidence. This duty is codified. Mark, let's
00:08:52 --> 00:08:57 read 20 CFR 404 . Okay. When the evidence
00:08:57 --> 00:09:00 in your case record is not sufficient for us
00:09:00 --> 00:09:03 to make a determination or decision. We will
00:09:03 --> 00:09:06 take the following actions. One, we will ask
00:09:06 --> 00:09:08 you to submit additional existing evidence. Two,
00:09:08 --> 00:09:10 we will seek additional existing evidence from
00:09:10 --> 00:09:13 your medical sources, employers, or other persons
00:09:13 --> 00:09:16 or organizations. Three, we may ask you to undergo
00:09:16 --> 00:09:20 a consultative examination at our expense. Four,
00:09:20 --> 00:09:23 we will develop your complete medical history
00:09:23 --> 00:09:25 for at least 12 months preceding the month in
00:09:25 --> 00:09:27 which you file your application unless there
00:09:27 --> 00:09:29 is a reason to believe that development of an
00:09:29 --> 00:09:32 earlier period is necessary. All right. So it
00:09:32 --> 00:09:35 says we will take the following actions, not
00:09:35 --> 00:09:38 may, not can. Also notice how this is written.
00:09:38 --> 00:09:42 A lot of regulations are written in pretty formal,
00:09:42 --> 00:09:45 dense legalese. This is designed as much as possible
00:09:45 --> 00:09:47 to be accessible to the people it's affecting.
00:09:47 --> 00:09:50 But it also says that the ALJ has an affirmative
00:09:50 --> 00:09:53 duty to develop a full and fair record. If there
00:09:53 --> 00:09:55 are gaps in the medical evidence, they're supposed
00:09:55 --> 00:09:57 to get the records. If your file is incomplete,
00:09:57 --> 00:10:00 they're supposed to. have you undergo whatever
00:10:00 --> 00:10:02 exams are needed at the government's expense.
00:10:02 --> 00:10:04 So the claimant who can't afford a specialist
00:10:04 --> 00:10:07 evaluation gets one anyway because the ALJ orders
00:10:07 --> 00:10:10 it. Yeah, so that's the design. And for a lot
00:10:10 --> 00:10:12 of people, it works. The ALJ obtains the records
00:10:12 --> 00:10:15 that the claimant might not have known existed
00:10:15 --> 00:10:18 or were needed. They order an exam that the claimant
00:10:18 --> 00:10:20 couldn't afford. They're developing evidence
00:10:20 --> 00:10:23 to prove the disability. But that same process
00:10:23 --> 00:10:26 can also go the other way. It absolutely can.
00:10:26 --> 00:10:28 And this is where I want to slow down because
00:10:28 --> 00:10:31 how this actually feels to the people in it matter.
00:10:31 --> 00:10:34 So we're going to try and work through a specific
00:10:34 --> 00:10:38 example. So we have Maria, and Maria has applied
00:10:38 --> 00:10:40 for disability based on chronic pain and depression.
00:10:41 --> 00:10:43 She's been seeing her primary care doctor, but
00:10:43 --> 00:10:45 there's no specialist documentation in the file.
00:10:45 --> 00:10:48 There's no psychiatric evaluation. There's no
00:10:48 --> 00:10:50 pain management records. There's been limited
00:10:50 --> 00:10:53 objective testing. So there's gaps in the evidence.
00:10:53 --> 00:10:56 The ALJ reviews the file before the hearing and
00:10:56 --> 00:10:59 sees those gaps. So the ALJ orders a consultative
00:10:59 --> 00:11:02 examination with a psychiatrist. OK, that sounds
00:11:02 --> 00:11:04 helpful. She couldn't afford to see the psychiatrist
00:11:04 --> 00:11:08 on her own. So maybe. Let's say she meets with
00:11:08 --> 00:11:10 the psychiatrist once. It might be 30 minutes.
00:11:10 --> 00:11:13 The doctor does a mental status exam, reviews
00:11:13 --> 00:11:15 her history, writes a report. That report goes
00:11:15 --> 00:11:18 in the file. And what if the report says her
00:11:18 --> 00:11:21 symptoms are mild? So then the ALJ, this exact
00:11:21 --> 00:11:24 same person who decided to order the exam, now
00:11:24 --> 00:11:27 has evidence that undermines her claim. Evidence
00:11:27 --> 00:11:29 that she never asked for. Evidence that was generated
00:11:29 --> 00:11:32 by a doctor she didn't choose. Evidence that
00:11:32 --> 00:11:34 was created in a 30 -minute appointment that
00:11:34 --> 00:11:36 was never going to capture what her own personal
00:11:36 --> 00:11:39 doctor had been seeing for years. And she's sitting
00:11:39 --> 00:11:40 in the hearing room, and this is the evidence
00:11:40 --> 00:11:43 the judge is looking at. Yes. Really try and
00:11:43 --> 00:11:45 think about this. You're Maria. You've been in
00:11:45 --> 00:11:47 pain for years. You've already been denied twice
00:11:47 --> 00:11:49 because that's what you need to do before you're
00:11:49 --> 00:11:51 even getting to the hearing. You finally have
00:11:51 --> 00:11:54 your hearing. You walk into the judge. This person
00:11:54 --> 00:11:56 who is supposed to be giving you a fair shot
00:11:56 --> 00:11:58 has already read everything. They ordered this
00:11:58 --> 00:12:00 exam you didn't want. And now they have a report
00:12:00 --> 00:12:02 in front of them that says you are not as bad
00:12:02 --> 00:12:05 off as you claim. The judge is going to question
00:12:05 --> 00:12:07 you about the inconsistencies between your testimony
00:12:07 --> 00:12:10 and this report that they ordered. That doesn't
00:12:10 --> 00:12:12 feel like a fair hearing. That feels like a cross
00:12:12 --> 00:12:14 -examination where the other side is also the
00:12:14 --> 00:12:17 judge. And here's the thing. This is absolutely
00:12:17 --> 00:12:20 not a legal violation. The ALJ is doing exactly
00:12:20 --> 00:12:22 what the regulations require, developing a full
00:12:22 --> 00:12:25 and fair record. That consultative exam is a
00:12:25 --> 00:12:28 standard tool. This questioning is part of the
00:12:28 --> 00:12:30 ALJ's role. Everything that happened to Maria
00:12:30 --> 00:12:34 is the system working as designed. Which is kind
00:12:34 --> 00:12:37 of the problem, right? Yeah. This isn't about
00:12:37 --> 00:12:39 bad actors. It's about a system where the structural
00:12:39 --> 00:12:42 incentives can create risks, even if everyone
00:12:42 --> 00:12:44 is trying to do their job as conscientiously
00:12:44 --> 00:12:47 as possible. What we're talking about here aren't
00:12:47 --> 00:12:50 legal violations. These are design risks. And
00:12:50 --> 00:12:52 they're the kind that emerge from how human cognition
00:12:52 --> 00:12:55 interacts with institutional structure. What
00:12:55 --> 00:12:57 exactly is the core risk here? So the biggest
00:12:57 --> 00:13:00 problem is confirmation bias. Once you've formed
00:13:00 --> 00:13:02 a hypothesis, you tend to seek out and weigh
00:13:02 --> 00:13:05 evidence that confirms it and to discount evidence
00:13:05 --> 00:13:08 that contradicts it. This isn't a character flaw.
00:13:08 --> 00:13:11 It's just how people's brains work. And we've
00:13:11 --> 00:13:13 got decades of cognitive science research saying
00:13:13 --> 00:13:16 this. Right. This is Fedor Dostoevsky's problem
00:13:16 --> 00:13:19 in Crime and Punishment, right? Is the inspector
00:13:19 --> 00:13:22 inherently biased against the defendant, right?
00:13:22 --> 00:13:25 And this kind of affects everyone, scientists,
00:13:25 --> 00:13:29 doctors, judges. Yeah. Imagine an ALJ who reviews
00:13:29 --> 00:13:32 Maria's files before the hearing. The prior denials
00:13:32 --> 00:13:34 are in there. The limited medical evidence is
00:13:34 --> 00:13:36 in there. The ALJ starts forming an impression.
00:13:36 --> 00:13:39 Maybe this person's condition isn't well supported.
00:13:40 --> 00:13:42 They don't have a conclusion yet. This is just
00:13:42 --> 00:13:44 an impression. And then the ALJ has to decide
00:13:44 --> 00:13:46 what additional evidence they want to develop.
00:13:47 --> 00:13:50 Exactly. Do they order a consultative exam to
00:13:50 --> 00:13:53 try and catch inconsistencies? Or when designed
00:13:53 --> 00:13:55 to explore the full extent of the disability?
00:13:55 --> 00:13:58 The regulation says develop a full and fair record,
00:13:58 --> 00:14:01 but full and fair in which direction? And the
00:14:01 --> 00:14:04 ALJ might not even be aware of the bias it's
00:14:04 --> 00:14:06 operating under. Right. That's the whole point
00:14:06 --> 00:14:08 of implicit bias. It is absolutely invisible
00:14:08 --> 00:14:11 to the person experiencing. The ALJ might sincerely
00:14:11 --> 00:14:14 believe they're being neutral while unconsciously
00:14:14 --> 00:14:16 steering this investigation towards confirming
00:14:16 --> 00:14:18 their preliminary view. And here's what I keep
00:14:18 --> 00:14:21 thinking. In the adversarial system, we have
00:14:21 --> 00:14:24 a check for this. If the prosecutor's theory
00:14:24 --> 00:14:27 is wrong, the defense attorney is there to expose
00:14:27 --> 00:14:29 it. If one side's evidence is weak, the other
00:14:29 --> 00:14:32 side tears it apart. The adversarial system has
00:14:32 --> 00:14:35 a built -in error correction mechanism. Right.
00:14:35 --> 00:14:37 And the inquisitorial model in Social Security
00:14:37 --> 00:14:39 doesn't have that same mechanism. There's no
00:14:39 --> 00:14:41 opposing counsel in the hearing room. The government
00:14:41 --> 00:14:43 doesn't send in a lawyer to argue against the
00:14:43 --> 00:14:46 claimant. It's the ALJ and the claimant. So if
00:14:46 --> 00:14:49 the ALJ's investigation goes sideways, there's
00:14:49 --> 00:14:51 nobody there to catch it in real time. Right.
00:14:51 --> 00:14:53 Assuming that this person isn't represented,
00:14:53 --> 00:14:56 there's nobody there who has the sophistication
00:14:56 --> 00:14:58 to understand what's actually happening. The
00:14:58 --> 00:15:00 claimant can object to what's going on, but they're
00:15:00 --> 00:15:02 objecting to the person who made that decision,
00:15:03 --> 00:15:06 who is also the person deciding their case. That
00:15:06 --> 00:15:07 is definitely not the same as having an independent
00:15:07 --> 00:15:10 advocate pushing back on it. And once you've
00:15:10 --> 00:15:12 built the record, it's hard not to believe it.
00:15:13 --> 00:15:15 Right. So this is again related to the ownership
00:15:15 --> 00:15:17 effect or endowment effect, but I like ownership.
00:15:17 --> 00:15:19 When you've invested effort in developing something
00:15:19 --> 00:15:23 like a theory, an argument, a body of evidence.
00:15:23 --> 00:15:26 a case file, you tend to value it more highly
00:15:26 --> 00:15:28 than if someone else had handed it to you. An
00:15:28 --> 00:15:30 ALJ who spent weeks building this record has
00:15:30 --> 00:15:32 a psychological investment in that record being
00:15:32 --> 00:15:35 right. It's not just confirmation bias. It's
00:15:35 --> 00:15:37 that the act of investigating creates its own
00:15:37 --> 00:15:40 momentum. And it can pound. Preliminary impressions
00:15:40 --> 00:15:42 shape the investigation. The investigation produces
00:15:42 --> 00:15:45 evidence. The evidence can then reinforce that
00:15:45 --> 00:15:47 impression. By the time you get to the hearing,
00:15:47 --> 00:15:49 the trajectory can be set. Not because anyone's
00:15:49 --> 00:15:51 intentionally doing anything wrong, but because
00:15:51 --> 00:15:54 that's how the cognitive loop works. There's
00:15:54 --> 00:15:56 also an appearance problem, right? Because separate
00:15:56 --> 00:16:01 from whether actual bias exists... Even if ALJs
00:16:01 --> 00:16:03 are perfectly fair in every single case, how
00:16:03 --> 00:16:05 does it feel to sit in that hearing room knowing
00:16:05 --> 00:16:07 that the judge has already reviewed everything
00:16:07 --> 00:16:09 and gathered more evidence and formed impressions?
00:16:10 --> 00:16:12 For a lot of claimants, it feels like the decision
00:16:12 --> 00:16:14 was made before they walked in the door. And
00:16:14 --> 00:16:17 perception of fairness matters for legitimacy.
00:16:17 --> 00:16:20 It does. If people don't trust the process, they
00:16:20 --> 00:16:22 don't trust the outcomes, even if the outcomes
00:16:22 --> 00:16:25 are correct. So this erosion of trust has real
00:16:25 --> 00:16:27 consequences for a system that's processing hundreds
00:16:27 --> 00:16:30 of thousands of cases a year. We can also ask,
00:16:30 --> 00:16:33 is this constitutional? Does combining this investigation
00:16:33 --> 00:16:35 and adjudication in the same person violate due
00:16:35 --> 00:16:38 process? I mean, based on everything we've talked
00:16:38 --> 00:16:42 about, it seems like it should. Right. And obviously,
00:16:42 --> 00:16:43 since this is in place, we're not going to be
00:16:43 --> 00:16:45 talking about courts that have said it does.
00:16:46 --> 00:16:48 Instead, the Supreme Court directly addressed
00:16:48 --> 00:16:53 this with Ruby Larkin in 1975. And here the case
00:16:53 --> 00:16:55 was talking about a Wisconsin medical licensing
00:16:55 --> 00:16:58 board that had acted as both... investigator
00:16:58 --> 00:17:02 and then a decider on whether to suspend a doctor's
00:17:02 --> 00:17:06 license. Here's the key passage. The contention
00:17:06 --> 00:17:09 that the combination of investigative and adjudicative
00:17:09 --> 00:17:11 functions necessarily creates an unconstitutional
00:17:11 --> 00:17:14 risk of bias in administrative adjudication has
00:17:14 --> 00:17:16 a much more difficult burden of persuasion to
00:17:16 --> 00:17:19 carry. It must overcome a presumption of honesty
00:17:19 --> 00:17:22 and integrity in those serving as adjudicators.
00:17:22 --> 00:17:26 And it must convince that under a realistic appraisal
00:17:26 --> 00:17:28 of psychological tendencies and human weakness,
00:17:28 --> 00:17:30 conferring investigative and adjudicated powers
00:17:30 --> 00:17:33 on the same individuals poses such a risk of
00:17:33 --> 00:17:36 actual bias or prejudgment that the practice
00:17:36 --> 00:17:39 must be forbidden if the guarantee of due process
00:17:39 --> 00:17:41 is to be adequately implemented. Must overcome
00:17:41 --> 00:17:44 a presumption of honesty and integrity. The court
00:17:44 --> 00:17:46 here is starting from the assumption that adjudicators
00:17:46 --> 00:17:49 are fair. And it's up to the challenger to try
00:17:49 --> 00:17:51 and show that combining these functions in this
00:17:51 --> 00:17:54 particular case pose such a risk of actual bias
00:17:54 --> 00:17:56 or prejudgment that the practice must be forbidden.
00:17:57 --> 00:18:00 That is a really high bar. Yeah, especially when
00:18:00 --> 00:18:02 we've just spent 10 minutes talking about how
00:18:02 --> 00:18:06 the bias is implicit, invisible even to the person
00:18:06 --> 00:18:09 experiencing. How would you even establish that?
00:18:09 --> 00:18:12 If the bias operates through investigation choices
00:18:12 --> 00:18:14 and evidence weighing rather than through overt
00:18:14 --> 00:18:16 statements of prejudice. How does the claimant
00:18:16 --> 00:18:19 actually prove this? That's the central tension.
00:18:19 --> 00:18:21 The court has set up this standard that requires
00:18:21 --> 00:18:24 proof of actual bias, but the whole nature of
00:18:24 --> 00:18:26 the problem is that the bias is generally structural,
00:18:26 --> 00:18:29 not individual. You're not going to find an ALJ
00:18:29 --> 00:18:31 who says, I decided against you because I investigated
00:18:31 --> 00:18:35 you and I didn't like it. I sided against you
00:18:35 --> 00:18:37 because I don't like you and I was in charge
00:18:37 --> 00:18:39 of the investigation. This bias is operating
00:18:39 --> 00:18:41 beneath the surface. The constitutional standards
00:18:41 --> 00:18:44 essentially prove something that by its nature
00:18:44 --> 00:18:47 can't be proven except for in the most extreme
00:18:47 --> 00:18:51 cases. Yes. So that's a fair characterization
00:18:51 --> 00:18:53 of the critics view. Defenders would say that
00:18:53 --> 00:18:55 this is pragmatic, that we can't restructure
00:18:55 --> 00:18:58 every government adjudication system based on
00:18:58 --> 00:19:01 theoretical risks. The question is whether the
00:19:01 --> 00:19:03 specific combination in a specific context creates
00:19:03 --> 00:19:07 an unacceptable risk of actual bias. And no one
00:19:07 --> 00:19:08 has successfully made that showing for Social
00:19:08 --> 00:19:11 Security. Okay. What would cross the line? The
00:19:11 --> 00:19:13 court suggested a few things. If somebody had
00:19:13 --> 00:19:16 a direct financial interest in the outcome, or
00:19:16 --> 00:19:18 if there was personal involvement that went beyond
00:19:18 --> 00:19:20 a standard investigation, or if there's situations
00:19:20 --> 00:19:22 where the adjudicator had essentially already
00:19:22 --> 00:19:25 committed to the problem, but the structure itself
00:19:25 --> 00:19:27 is investigating and judging, that's permissible.
00:19:28 --> 00:19:30 So courts have consistently upheld the system.
00:19:30 --> 00:19:33 And the standard argument is that, again, the
00:19:33 --> 00:19:36 reason this was created, the ALJ's duty to develop
00:19:36 --> 00:19:38 the record helps the claimants more than it hurts
00:19:38 --> 00:19:41 them. And any alternative would be worse. Do
00:19:41 --> 00:19:44 you buy that? So I think it's pretty complicated.
00:19:44 --> 00:19:49 For some claimants who are unsophisticated, unrepresented
00:19:49 --> 00:19:51 and unable to navigate the system, having the
00:19:51 --> 00:19:53 ALJ develop evidence probably really does help.
00:19:54 --> 00:19:56 The ALJ really is ordering exams the claimant
00:19:56 --> 00:19:58 couldn't have afforded that. proves whatever
00:19:58 --> 00:20:01 they were trying to argue. It gets the records
00:20:01 --> 00:20:02 that claim it didn't know existed. That's real.
00:20:03 --> 00:20:06 OK, but what about in situations like with the
00:20:06 --> 00:20:08 current administration? Most of what we're trying
00:20:08 --> 00:20:12 to talk about is neutral before things started
00:20:12 --> 00:20:16 being attacked. The administrative state is absolutely
00:20:16 --> 00:20:19 under attack right now. That's kind of the reason
00:20:19 --> 00:20:23 for the podcast. But. If the system is designed
00:20:23 --> 00:20:25 to work as intended when people have resources,
00:20:25 --> 00:20:28 and even then we're talking about problems with
00:20:28 --> 00:20:29 the system, which is what we've been talking
00:20:29 --> 00:20:32 about this whole time, is the system working
00:20:32 --> 00:20:35 as it was intended to work. When you take away
00:20:35 --> 00:20:37 those resources, we have a complete and utter
00:20:37 --> 00:20:39 failure. We're going to be talking about the
00:20:39 --> 00:20:41 importance of due process in the next episode,
00:20:41 --> 00:20:44 and due process is critical. Because that is
00:20:44 --> 00:20:46 how we ensure that the actual right decision
00:20:46 --> 00:20:49 is reached. The mere fact that somebody has been
00:20:49 --> 00:20:50 picked up doesn't mean that they were correctly
00:20:50 --> 00:20:53 picked up. And if they don't have the opportunity
00:20:53 --> 00:20:57 to demonstrate that, everybody is at risk. So
00:20:57 --> 00:21:01 it is absolutely a huge problem right now. But
00:21:01 --> 00:21:03 even if we're not talking about the current administration,
00:21:04 --> 00:21:06 we're talking about before things started to
00:21:06 --> 00:21:10 get really crazy, in general, Just the investigation
00:21:10 --> 00:21:13 itself can become a way that the ALJ develops
00:21:13 --> 00:21:15 evidence undermining the claim and for which
00:21:15 --> 00:21:17 the claimant has no meaningful way to push back.
00:21:18 --> 00:21:21 So the structure cuts both ways. Again, when
00:21:21 --> 00:21:23 it's fully funded, when it's operating as intended,
00:21:23 --> 00:21:27 when we're dealing with a normal caseload. And
00:21:27 --> 00:21:30 the particular way it's going to cut in any given
00:21:30 --> 00:21:32 case depends on factors that the claimant can't
00:21:32 --> 00:21:35 control. And the constitutional standard says
00:21:35 --> 00:21:38 that's fine as long as you can't prove actual
00:21:38 --> 00:21:41 bias in your specific case. Yes, which is pretty
00:21:41 --> 00:21:44 much impossible to prove if the bias is structural.
00:21:45 --> 00:21:48 Okay. I want to make sure we're being fair to
00:21:48 --> 00:21:50 the system here because there's a version of
00:21:50 --> 00:21:53 this story where the inquisitorial model is actually
00:21:53 --> 00:21:56 the more humane option. All right. Okay. Picture
00:21:56 --> 00:22:00 the alternative. Fully adversarial system. The
00:22:00 --> 00:22:02 government sends a lawyer to argue against every
00:22:02 --> 00:22:05 disability claimant. You're already sick, already
00:22:05 --> 00:22:08 struggling financially, and now there's a government
00:22:08 --> 00:22:11 attorney whose job is to defeat your claim. That's
00:22:11 --> 00:22:13 what the adversarial system would look like in
00:22:13 --> 00:22:15 practice for someone applying for disability.
00:22:15 --> 00:22:18 Right, and that's not hypothetical. A lot of
00:22:18 --> 00:22:20 other benefit systems work closer to that model,
00:22:20 --> 00:22:23 and claimants often do fare worse in those kinds
00:22:23 --> 00:22:25 of systems without representation. So the people
00:22:25 --> 00:22:28 who designed Social Security Disability looked
00:22:28 --> 00:22:30 at both options and said the inquisitorial model
00:22:30 --> 00:22:34 where the judge helps build your case is actually
00:22:34 --> 00:22:37 kinder to claimants than forcing them into a
00:22:37 --> 00:22:39 fight that they can't win. Yeah, it certainly
00:22:39 --> 00:22:42 seems like that's why things have been structured
00:22:42 --> 00:22:45 the way they have. And it is definitely important
00:22:45 --> 00:22:47 to take that seriously. The alternative isn't
00:22:47 --> 00:22:50 some platonic ideal of perfect fairness. The
00:22:50 --> 00:22:52 alternative has a different set of problems.
00:22:52 --> 00:22:54 An adversarial system where government always
00:22:54 --> 00:22:57 has better lawyers. Or almost always. So, yeah,
00:22:58 --> 00:23:00 you're choosing between a system where the judge
00:23:00 --> 00:23:02 investigates you with all the cognitive risks
00:23:02 --> 00:23:05 we talked about there and a system where a government
00:23:05 --> 00:23:07 attorney argues against you with all the resource
00:23:07 --> 00:23:10 balance that entails. Neither is perfect. The
00:23:10 --> 00:23:12 question is which kind of imperfections you want
00:23:12 --> 00:23:15 to live with. And Congress chose the investigative
00:23:15 --> 00:23:18 judge. For this particular context. But, again,
00:23:18 --> 00:23:20 it's worth noting that they made different choices
00:23:20 --> 00:23:22 for different systems. For immigration proceedings,
00:23:23 --> 00:23:25 they are adversarial. There's a government attorney.
00:23:25 --> 00:23:29 For SEC enforcement, those actions are also adversarial.
00:23:29 --> 00:23:32 Tax court's adversarial. Social security disability
00:23:32 --> 00:23:35 really is the outlier here, and it's the outlier
00:23:35 --> 00:23:37 because the claimant population was seen as especially
00:23:37 --> 00:23:40 vulnerable. So the design reflects a judgment
00:23:40 --> 00:23:43 about who the claimants are. Yes, it does, and
00:23:43 --> 00:23:45 about the best way to help them. Okay, Gwen,
00:23:45 --> 00:23:48 you know what keeps sticking with me? The word
00:23:48 --> 00:23:50 hearing. When a claimant gets a letter saying
00:23:50 --> 00:23:52 you have a right to a hearing, they think that
00:23:52 --> 00:23:54 means someone's going to sit down and listen
00:23:54 --> 00:23:56 to their side. They don't expect the person listening
00:23:56 --> 00:23:59 to have already ordered exams, formed impressions,
00:23:59 --> 00:24:02 based on the record they helped build. Yeah,
00:24:02 --> 00:24:04 people can feel blindsided here. Not because
00:24:04 --> 00:24:06 the system is trying to trick them, but because
00:24:06 --> 00:24:07 nobody explained what they were walking into.
00:24:08 --> 00:24:11 Okay, then so what comes next? Because so far,
00:24:11 --> 00:24:14 we've been talking about what agencies choose
00:24:14 --> 00:24:16 to do. At some point, the Constitution has to
00:24:16 --> 00:24:19 start telling them what they must do. And that's
00:24:19 --> 00:24:21 exactly where we're heading. As I mentioned earlier,
00:24:22 --> 00:24:24 next episode, we're talking about why the Constitution
00:24:24 --> 00:24:26 requires the government to hear from you before
00:24:26 --> 00:24:29 it takes something away and why that requirement
00:24:29 --> 00:24:35 is thinner than most people think. So that does
00:24:35 --> 00:24:36 it for today's episode on administrative remedies.
00:24:37 --> 00:24:39 Thank you for joining us today. Please, if you
00:24:39 --> 00:24:42 enjoy this podcast and enjoy this episode, give
00:24:42 --> 00:24:46 us a like on Spotify, iTunes, or whatever platform
00:24:46 --> 00:24:48 you're listening on. And be sure to tune in next
00:24:48 --> 00:24:51 time where we'll continue to dive into the contours
00:24:51 --> 00:24:53 of administrative law. Because remember, you
00:24:53 --> 00:24:54 can't fix what you don't understand.