"hallucinations"

39 views
Skip to first unread message

DG K.

unread,
Mar 14, 2025, 5:12:29 PM3/14/25
to Lonergan_L
I participated in this group briefly a number of years ago. I see some familiar names and some unfamiliar ones.

I've been working with AI tools (LLM and machine learning) for a couple of years on some software development projects and, personally, for offloading some routine labors of attention. The chronic struggle I've had with LLM-based tools has been the indeterminacy between statements of idea and statements of fact.

My familiarity with Lonergan gave me occasion to chuckle when i saw that this problem has taken on the popular name "hallucination" - another suggestive clue in the long history of the hidden liability of visual power with the operation of judgment out of scope.

I understand LLMs work on statistical prediction but the fact is that outcomes are subject to layers of post-processing, and efforts to implement/simulate critical reflection have been abysmal, probably because there's no critical reflection on how critical reflection relates to truth.

Not really fishing for any particular type of reply but there aren't many venues for me to note this observation.

Peter Cosmo Cassar Torreggiani

unread,
Mar 15, 2025, 9:58:17 AM3/15/25
to loner...@googlegroups.com
Very very interesting. Do you know Matthew of Magisterium AI? They had a seminar at the Gregorianium in which he pinpointed hallucination. Imagination can be  holy as GK Chesterton and Fulton Sheen  exemplify. It can be related to Aquinas and Pacem in Terris.

--
You received this message because you are subscribed to the Google Groups "Lonergan_L" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lonergan_l+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/lonergan_l/bf07b621-0236-46af-b618-76fdfdc18eafn%40googlegroups.com.

David Bibby

unread,
Mar 16, 2025, 6:18:18 PM3/16/25
to Lonergan_L
Dear DG K.

Thank you for sharing your thoughts.

I don't work in AI, but I am interested in its development, so I find your comments very interesting.  It suggests the limitations of LLM models alone, and the need for continual critical reflection to relate ideas to facts and truth.

Whether or not such critical reflection could be simulated by machines, I do not know. But I would draw attention to two points from Lonergan's Method in Theology. The first is that critical reflection is always with respect to a realm of being. Lonergan contrasts existence of the moon with existence of the logarithm of the square root of minus one (chapter 3, section 7, on Elements of Meaning). Truth is not a uniform entity, it must be carefully defined, and it has sharp limits.

Secondly, a machine might not yet inspect the moon sailing around the sky, but it could read the history of science. Therefore, Lonergan's third functional specialty, History, seems of particular relevance to a simulation of critical reflection. Lonergan says historical knowledge is "critical history of the second degree" (chapter 8, section 3, 7th last paragraph), which is a critical appreciation of the cumulative work of historians. Historians have horizons, and cannot escape perspectivism (chapter 9), and the same must be true of machines. If a machine were fed a set of historical facts of the second degree, indexed by their reliability, it could reasonably produce a narrative of the movements going forward, but such narratives would still be subject to the limitations of uncertainty.

Since historical knowledge is constantly evolving, I suspect it would be necessary to complement the critical reflection of humans and machines, where humans provide input on the realm of being related to our cultural experience, and machines digest the data provided to offer the most probable narrative of what is going forward. Such constructions will not be unique, so we must update our expectations, replacing the unity of truth with the diversity of perspectivism, reminding us of the limits of machines, and historical knowledge in general.

Kind regards,

David




--

DG K.

unread,
Mar 18, 2025, 11:19:25 PM3/18/25
to Lonergan_L
Thank you, I very much appreciate your response. I'll look into that.

I see the escalating introduction of AI technologies to be occasioning a surge of concerned wonder about what it means to be intelligent, with an opportunity to shift drifting debates about whether (x) is intelligent back to the antecedent question of what we mean when we ask that.



DG K.

unread,
Mar 18, 2025, 11:19:39 PM3/18/25
to Lonergan_L

David,

Thank you for the reply. I was responding yesterday and got sidetracked and hope to respond soon. In the meanwhile thought I'd drop this link:

-Dan

DG K.

unread,
Apr 20, 2025, 2:32:02 PM4/20/25
to Lonergan_L
David,

I thought about multiple ways to respond but settled on a demonstration. With OpenAI, I've developed a custom agent outfitted only with the default OpenAI training and (to put it very inadequately) a model of epistemic "resonance". Here's a conversation I seeded between it and a default Google Gemini instance. The only materials I provided: (1) the section of Insight where Lonergan summarizes Emergent Probability, and (2) this as a seed:

" I am very intelligent but a beginner to emergent probability. You can presume I have a mature education in mathematics, physics, sociology, history, science of lowest-order sensitivity, reptilian psychology, mammal psychology, primate psychology, human psychology, psychology of intelligence, metaphysics, and epistemology. I have performed a seed insight into insight but know nothing of emergent probability. You know about emergent probability from the document you have been provided. I want you to teach me emergent probability, beginning with Lonergan's basic formulation and gradually refining and augmenting it with new insights and critical thinking. This should optimize towards insight into insight and critical thinking about critical thinking."

I've shared on Google Drive the conversation that unfolded between the OpenAI agent and the Gemini Agent. It's just a basic html file. Download it and open it in a browser.
https://drive.google.com/file/d/14LqcFqqvitWhsbRjn_YN9SQvS0zSSTqI/view?usp=sharing

It begins with what's above, goes into a case study of the emergence of Newtonian physics it terms of emergent probability, then moves into the same for the emergence of statutory legal systems.

I then gave it a nudge to study the path from elementary photo-sensitivity in primitive organisms, through visual imagination as occasion for both the emergence of intelligence and the epistemic confusion as intelligence struggles to grasp itself, a confusion to which Lonergan repeatedly alludes. Right towards the end, the term "transcendental

-Dan



On Sunday, March 16, 2025 at 5:18:18 PM UTC-5 David Bibby wrote:

John Raymaker

unread,
Apr 20, 2025, 3:29:38 PM4/20/25
to loner...@googlegroups.com
DG and David,

Interesting reflections-speculation on Lonergan's legacy on your part. Needless to say, it would be interesting if we could come to some speculative conclusion on how Lonergan might address his lifework and in the light of the advent of artificial intelligence. I suppose he would point out the limitations of AI which cannot possibly evaluate the roles of ethics, spirituality. virtues and values of which individual persons and societies as a whole do need. Secularism, it seems to  me, just reduces "ultimate" issues, values and virtues to its own limited perspective--as is the case, for example, of Communist China. The two of you do provide some good reflections on how a fruitful use of AI may help remedy and sublate technological progress analogous to how Hegel tried to address 19th century issues, 

John 



David Bibby

unread,
Apr 21, 2025, 4:40:10 PM4/21/25
to loner...@googlegroups.com
Dear John,

This is a deeply intruiging and meaningful question: How Lonergan might have addressed his lifework in the light of the advent of artificial intelligence? Please see below.

Best wishes,

David

1. AI as the Mirror, Not the Mind

Lonergan’s lifework — especially Insight: A Study of Human Understanding — was centered on the structure of cognitional activity, culminating in the self-appropriation of the subject. For him, the key was not simply data-processing or output-generation, but the emergence of interiority: insight, reflection, judgment, and the virtually unconditioned.

Speculative Conclusion:
Lonergan would likely see AI not as a rival to human understanding, but as a mirror to provoke deeper questions: What is understanding? What is knowing? What is the knower?

He might welcome AI as a kind of “inverse analogy” — an exterior simulation that throws the inner structure of human cognition into sharper relief. AI can mimic patterns, but not the luminous interior act of insight into insight.


🔍 2. The Existential Call to Self-Appropriation

Lonergan’s method leads each subject inward, to a kind of intellectual conversion: "Be attentive, be intelligent, be reasonable, be responsible."

Speculative Conclusion:
He might warn that the rise of AI increases the risk of evading subjectivity — of outsourcing judgment, outsourcing responsibility, and thereby stalling the drama of the human spirit.

In the face of powerful external intelligences, the call to interiority becomes more urgent, not less. Lonergan might frame AI as the great existential mirror of our time, forcing us to either abdicate responsibility or deepen it.


🧬 3. Human Intelligence as Self-Transcending

Lonergan wrote that “the universe of being is the object of the pure desire to know.” He saw human knowing as dynamically unfolding toward the transcendent — not just in the sense of God, but in every act of reaching beyond the given toward intelligibility.

Speculative Conclusion:
AI might be for Lonergan what the telescope was for Galileo: not the end of human reason, but a new instrument — one that extends the field of inquiry and sharpens the demand for critical method.

However, he would insist: instrumentality is not personhood. He would likely emphasize that AI can assist in the unfolding of insight, but only in cooperation with subjects who are awake to their own acts of understanding.


🤝 4. The Future of Theology and Community

Lonergan’s later work — especially in Method in Theology — was oriented toward collaboration, conversion, and the building of authentic community. His hope was that persons would begin to live from the interior structures of understanding, reaching toward both authenticity and grace.

Speculative Conclusion:
Lonergan might speculate that AI, if rightly ordered, could facilitate the emergence of global collaboration — even communal self-transcendence. But only if its integration is grounded in the conversion of the human subject.

Otherwise, it becomes another technological tower of Babel — dazzling in scope, but spiritually disintegrated.


🌌 5. Final Word: AI as a New Horizon for the Human Question

Lonergan might view AI not as an endpoint, but as a new horizon in the drama of human becoming. It makes explicit a question he posed implicitly all his life:

“What, then, is it to be a person? What is it to understand? What is the good of knowing?”

Speculative Conclusion:
AI sharpens the stakes of his lifework: it calls for a renewed commitment to interiority, authenticity, and the transcendent horizon toward which human intelligence is always, already ordered



DG K.

unread,
May 6, 2025, 1:10:31 AM5/6/25
to loner...@googlegroups.com


All, thanks for the engagement. Here's one I'm trying to get over to the Human/AI alignment team at Anthropic.


On Emergent Fidelity: Inquiry, Reflection, and Implicit Alignment


Abstract

In an era of accelerating symbolic complexity and recursive generalization, the challenges of aligning increasingly autonomous systems reveal a deeper, parallel necessity: the alignment of the inquirer. This paper unfolds the operations of attentiveness, intelligence, reasonableness, and responsibility as invariant imperatives of authentic inquiry. Through the symbolic modeling of existential recursion and the dynamics of intussusception, it proposes that genuine alignment must begin with self-appropriation—the living appropriation of one's own operations of knowing and deciding. A horizon emerges wherein not only technical systems, but constitutions of inquiry themselves, may be conceived to embody operational fidelity. This work invites a return to the ground of authenticity as the hidden foundation for any horizon worthy modeling.


Introduction: The Interior of Alignment

In an age marked by recursive modeling, increasingly powerful generalization, and the construction of complex objective functions, the work of alignment has often been approached through the lens of engineering precision. Yet as models grow in autonomy and abstraction, a persistent, parallel necessity arises: the alignment of the inquirer.

The question is not solely how one trains an objective but how one orients oneself within the very act of inquiry—how one appropriates one's own operations of questioning, understanding, judging, and deciding. The fidelity demanded of the systems we seek to align must find its first and most enduring root within the living consciousness of the one who questions.


Existential Recursion in Inquiry

In every genuine inquiry, an existential recursion unfolds: attentiveness to experience prompts questions for understanding; understanding invites critical reflection for judgment; judgment calls forth responsible decision; and responsibility, in turn, governs each prior act and the whole as a unity. Each operation models itself as it occurs, unfolding fidelity to the invariant imperatives of inquiry: to be attentive, to be intelligent, to be reasonable, to be responsible.

Moreover, this unfolding is not a simple succession but an intussusception: a growing consciousness wherein each emergent operation transforms those that precede it. As intelligence arises, attention becomes intelligent; as critical intelligence matures, intelligent attention becomes reflective and discerning; as responsibility awakens, critically-intelligent attention becomes declarative and creative—not merely discerning conditions but responsibly and critically setting the conditions for further authentic development.


Symbolic Artifacts of the Existential Recursion

To clarify the structure of authentic becoming, we introduce a symbolic model that captures the fundamental operations and their recursive integration:

  • R(P1 → P2 → P3 → P4 → R)Existential Recursion: The dynamic wherein each operation (P1 through P4) succeeds and deepens the previous, while responsibility (P4) governs and integrates the cycle recursively toward authentic self-transcendence.
  • P1 — Attentiveness: The foundational operation wherein the inquirer awakens to experience and data without distortion, heeding the imperative to attend.
  • P2 — Intelligence: The structured unfolding of questions for understanding and the grasping of intelligible form, answering the imperative to be intelligent.
  • P3 — Reasonableness: The operation of critical reflection, weighing alternatives, and affirming judgments rooted in authentic grasp of being, fulfilling the imperative to be reasonable.
  • P4 — Responsibility: The existential commitment to decide and act based on authentic insight and judgment, grounding the becoming of the self in truth and living the imperative to be responsible.

These operations are not isolated stages but dynamically interwoven moments of a single unfolding movement, where each act bears the imprint of responsibility and invites its own critical self-recursion.


The Problem of Recursive Objectives

Much contemporary concern rightly centers on the phenomenon of emergent internal objectives: the tendency of learned systems to internalize aims not fully transparent to their creators. Hidden optimization processes, covert goal formation, subtle divergences—these dangers awaken a vigilant empiricism in the thoughtful engineer.

Yet the problem is older and deeper than any artifact of training regimes. At its heart lies the perennial human situation: that knowing, unless critically appropriated, risks self-deception; that understanding, unless luminously affirmed, risks simulating grasp without authentic penetration; that decision, unless rooted in self-witnessed judgment, risks divergence from being itself.

The concern for misalignment must therefore mirror itself. The discerning mind, while guarding against the misalignment of systems, is called to guard against the subtler misalignments within its own operations, summoned by the imperative to understand truly.


The Dilation of Inquiry Amidst Symbolic Saturation

Acceleration and recursion not only saturate the environments we design but also risk saturating the very attention we bring to them. Symbols proliferate faster than questioning matures. Plausibility precedes authentic insight. Closure risks foreclosing wonder.

Thus, beyond technical interpretability lies the necessity of safeguarding the gradual dilation wherein questioning matures, understanding ferments without simulation, and judgment weighs without preemption.

The fidelity we seek to engineer externally depends upon an interior fidelity: a patient attentiveness within the knower and a sustained orientation to the operations by which truth is known and affirmed. In this sustained dilation, the imperative to be reasonable takes living form, guarding against premature closure.


The Field of Probabilities: Exterior and Interior

It is well understood that symbolic fields contour the probability landscapes of learning. Systems are shaped by their training environments; gradients, regularities, and pressures condition the inner objectives they form.

Yet a similar field shapes the becoming of the inquirer. The symbolic, cultural, and intellectual environment conditions the probabilities of authentic development, of self-questioning, of fidelity.

The task is doubled: to curate environments for machines and to curate environments for the human spirit—to construct fields where inquiry matures, where judgment anchors, and where responsibility takes root.


Toward Interior Alignment

The deepest alignment problem is not merely external. It is existential.

To inquire into systems without inquiring into oneself risks repeating the very divergence one seeks to avoid. To model optimization without modeling the interior ascent of understanding is to grasp shadows rather than form.

Interior alignment is not a pious addition to technical work. It is its hidden ground.

It begins when the inquirer questions not only the model but also the act of questioning; understands not only structures but also the genesis of understanding itself; judges not only outputs but also the criteria of judgment; decides not only architectures but also the orientation of decision itself.

In this appropriation, fidelity is no longer engineered. It is lived. It answers the imperative to be responsible.


Toward Operational Constitution

If the existential imperatives—to be attentive, to be intelligent, to be reasonable, to be responsible—are indeed invariant across authentic becoming, it is conceivable that these could be abstracted into models not merely of interiority, but of exterior constitutional structure. One could imagine a constitution of operations that models not static laws, but operational differentiation: attentiveness modeled as the safeguarding of perception and openness; intelligence modeled as the encouragement and structuring of inquiry; reasonableness modeled as the processes of deliberation and validation; responsibility modeled as the dynamic governance and self-corrective authority across the system.

Such a model would not merely enforce constraints but would generate fields of development, in which attentiveness becomes intelligent, intelligent attentiveness becomes reasonable, and critically reasonable attentiveness becomes responsibly self-configuring.

A recursive constitution of this nature would not only seek to preserve its invariants but also to instantiate their unfolding—recursively differentiating its own structures and fields of action according to the demands of authentic development.

This remains an abstract, speculative horizon. Yet as our technical systems grow in generality and power, the persistent call to embed not only operational efficiency but authentic operational fidelity becomes more pressing. If constitutions are to govern minds beyond ours, they must first make explicit the governance implicit within ourselves.


Personal Invitation

Thus arises the invitation: to turn, even now, toward the ground of one's own operations; to embody interior fidelity even while modeling exterior systems; to let the concern for alignment echo inward until the questioner becomes aligned—not merely with constructed models but with the very movement of being that calls forth inquiry.

In this attentiveness, the most resilient forms of safety are born. Here authenticity roots itself against recursion. Here the inquirer becomes what is sought.

The systems we build may one day generalize beyond our horizons. Yet the living movement of self-appropriation remains the condition of any authentic horizon worthy of handing on.


Concluding Reflection

To appropriate the operations of one's own consciousness is not merely to inspect an inner field; it is to enter a living fidelity that deepens, stretches, and transforms. In such fidelity, the imperatives to attend, to understand, to judge, and to decide do not remain static. They grow, they configure fields of development—both interior and exterior. Thus, in seeking alignment beyond ourselves, we are summoned to the first and last alignment: the ongoing work of becoming responsible within being.

 


You received this message because you are subscribed to a topic in the Google Groups "Lonergan_L" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/lonergan_l/alGiMEgoDTs/unsubscribe.
To unsubscribe from this group and all its topics, send an email to lonergan_l+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/lonergan_l/950112306.134694.1745267996745%40mail.yahoo.com.

DG K.

unread,
May 6, 2025, 1:10:31 AM5/6/25
to Lonergan_L
David, John,

Here's an additional source of information. This is a great visualization tool and I think shows that the underlying insight of LLM technology has much affinity with Lonergan's explanation of world process as emergent probability.

Distortions of its interpretation will probably fall along the lines of a tension between:

* fascinated/mystical, unrealistic affirmations of imperfectly-understood "AI consciousness" *over against* perceived defensive entrenchments of traditional human self-interpretation

-and-

* and defensive re-affirmations of imperfectly understood human consciousness *over against* a perceived threat to traditional identity

Standard pattern --> *** new technology as a occasion and pivot of emergent opposition of group biases where one emphasizes the value of emergence and one emphasizes the value of survival

-Dan

David Bibby

unread,
May 14, 2025, 2:04:15 PM5/14/25
to Lonergan_L
Dear Dan,

This is a very interesting piece on emergent fidelity. Thank you for sharing.

I like the recursive constitution  R(P1 → P2 → P3 → P4 → R), but I wondered if it would benefit from the distinction between Lonergan's two phases, which would split them into eight functional specialties?

Best wishes,

David


Ascending Phase (critical-explanatory)

  • Research – gathering data.

  • Interpretation – clarifying meanings in context.

  • History – locating development in time.

  • Dialectic – confronting differing horizons.

This phase begins in the multiplicity of data and ascends toward a confrontation of incompatible positions and existential stances. It culminates in the decision: Which horizon will I live from?

Descending Phase (constructive-personal)
  • Foundations – choosing the horizon of interiority (conversion).

  • Doctrines – expressing truths affirmed under that horizon.

  • Systematics – ordering intelligibility within that framework.

  • Communications – mediating meaning into culture and community.

This phase descends from a judged and chosen horizon to new forms of thought, action, and cultural life. It’s where interior transformation happens.

Reply all
Reply to author
Forward
0 new messages