← Resources & Sources

brand/foundational/sources/bach.md

Sources: Joscha Bach

Contents
  1. Books
  2. Papers and essays
  3. Talks and interviews
  4. About him
  5. Could not verify

Joscha Bach

Photo: Joi Ito, Cambridge, MA, 11 Nov 2013, CC BY 2.0, via Wikimedia Commons. Cropped square.

Joscha Bach (b. 1973, Weimar) is a German cognitive scientist and AI researcher. He built the MicroPsi cognitive architecture on Dietrich Dörner's Psi theory. He argues that minds, selves and consciousness are models that a computational system builds of its world and of itself. This page lists his books, papers, talks and interviews, and where each can be read, watched or heard. Citations were checked against Crossref, arXiv, the publisher record, the event or podcast record, or the work itself. Compiled 2026-10-20.

Books

  • Bach, J. (2009). Principles of Synthetic Intelligence. PSI: An Architecture of Motivated Cognition. Oxford University Press (Oxford Series on Cognitive Models and Architectures). doi:10.1093/acprof:oso/9780195370676.001.0001. It sets out Dörner's Psi theory and then Bach's MicroPsi implementation of it. A mind here is a motivated agent. Physiological, social and cognitive needs drive what it does, and emotion emerges from how those needs modulate perception, memory and action selection. The chapters run from "Machines to explain the mind" through representation, behaviour control, language and Dörner's own agent, to the MicroPsi architecture and framework. Chapter 7 criticises the original Psi model, including how it handles symbol grounding. In copyright. The publisher's copy (Oxford Academic) is paywalled.
  • Bach, J. (2007). Principles of Synthetic Intelligence. Building Blocks for an Architecture of Motivated Cognition. PhD thesis, Institute of Cognitive Science, Universität Osnabrück. It is the dissertation the 2009 book grew out of. Open: author's draft, dated 30 March 2007, on cognitive-ai.com, the MicroPsi project site (7.4 MB). Sources disagree on the degree year (see Could not verify).

There are no local copies. The Internet Archive's lending copy of the 2009 book is a third-party copy and is not listed as a source.

Papers and essays

WorkFull textLocal copy
Bach, J. (2003). "The MicroPsi Agent Architecture". Proc. ICCM-5, Bamberg, 15–20. The first account of MicroPsi: agents built from node nets that implement Dörner's theory of motivation, emotion and action regulation.author PDF, cognitive-ai.comnone
Bach, J., & Vuine, R. (2003). "Designing Agents with MicroPsi Node Nets". KI 2003, LNAI 2821, 164–178. doi:10.1007/978-3-540-39451-8_13. Uses one network formalism, node nets, for control structures, plans, sensory schemas and neural learning.Springer (paywalled)none
Bach, J. (2005). "Representations for a Complex World: Combining Distributed and Localist Representations for Learning and Planning". In S. Wermter & G. Palm (Eds.), Biomimetic Neural Learning for Intelligent Robots, LNCS, 265–280. doi:10.1007/11521082_16. The same MicroPsi hierarchies work both as trainable neural networks and as symbolic plans and control structures for agents in a simulated world.author PDF, cognitive-ai.comnone
Bach, J., Bauer, C., & Vuine, R. (2007). "MicroPsi: Contributions to a Broad Architecture of Cognition". KI 2006: Advances in Artificial Intelligence, LNCS, 7–18. doi:10.1007/978-3-540-69912-5_2. Recasts the Psi theory of human action regulation as a cognitive modelling framework. Also gives an algorithm that matches perceptual input against hierarchical representations.author PDF, cognitive-ai.comnone
Bach, J. (2008). "Seven Principles of Synthetic Intelligence". Proc. First Conference on Artificial General Intelligence (AGI-08), Memphis, 63–74. ACM record 10.5555/1566174.1566182. Argues that AI dropped its original goal of understanding the mind, and names seven lessons for returning to it. They include building whole functionalist architectures, building autonomous systems with their own motivation, and not expecting intelligence to emerge by itself.author PDF, cognitive-ai.comnone
Bach, J. (2011). "A Motivational System for Cognitive AI". Artificial General Intelligence (AGI 2011), LNCS, 232–242. doi:10.1007/978-3-642-22887-2_24. A motivational system for cognitive agents, built from a finite set of drives that stand for the system's needs.Springer (paywalled)none
Adams, S. S., Arel, I., Bach, J., et al. (2012). "Mapping the Landscape of Human-Level Artificial General Intelligence". AI Magazine 33(1), 25–41. doi:10.1609/aimag.v33i1.2322. Twelve authors. A roadmap for AGI research: the capabilities a general intelligence needs, how to choose tasks and environments to test it, and seven milestone scenarios.AAAI OJS (open)none
Bach, J., & Verdicchio, M. (2012). "What Kind of Machine is the Mind?". Turing-100, EPiC Series in Computing 10, 16–19. doi:10.29007/k6f4. Many computational models of mind carry an unstated strong physicalism. The paper separates computationalism as an epistemological position from physical realisation, and treats symbol manipulation as constrained computation.EasyChair (open)none
Bach, J. (2012). "A Framework for Emergent Emotions, Based on Motivation and Cognitive Modulators". International Journal of Synthetic Emotions 3(1), 43–63. doi:10.4018/jse.2012010104. Emotions are not a separate module. They emerge from motivation and from the modulation of cognition, and the same terms can describe personality traits.IGI Global (paywalled)none
Bach, J. (2012). "Modeling Motivation and the Emergence of Affect in a Cognitive Agent". In P. Wang & B. Goertzel (Eds.), Theoretical Foundations of Artificial General Intelligence, Atlantis Press, 241–262. doi:10.2991/978-94-91216-62-6_13. A longer account of the same model: motivation as the base from which affect arises.no open copy foundnone
Bach, J. (2012). "MicroPsi 2: The Next Generation of the MicroPsi Framework". Artificial General Intelligence (AGI 2012), LNCS, 11–20. doi:10.1007/978-3-642-35506-6_2. A ground-up reimplementation of MicroPsi as a framework for agents built from neuro-symbolic spreading-activation networks. Code: github.com/joschabach/micropsi2.Springer (paywalled)none
Bach, J. (2015). "Modeling Motivation in MicroPsi 2". Artificial General Intelligence (AGI 2015), LNCS, 3–13. doi:10.1007/978-3-319-21365-1_1. The current motivation model, covering cognitive, social and physiological needs, and how it accounts for individual differences and personality.Springer (paywalled)none
Bach, J., & Gallagher, K. (2018). "Request Confirmation Networks in MicroPsi 2". Artificial General Intelligence (AGI 2018), LNCS, 12–20. doi:10.1007/978-3-319-97676-1_2. A network scheme in MicroPsi 2 for running hierarchical scripts and perceptual hypotheses, in which nodes request and confirm one another.Springer (paywalled)none
Bach, J. (2018). "The Cortical Conductor Theory: Towards Addressing Consciousness in AI Models". Biologically Inspired Cognitive Architectures 2018, Advances in Intelligent Systems and Computing, 16–26. doi:10.1007/978-3-319-99316-4_3. Sole author. A functionalist sketch of phenomenal consciousness: a "conductor" process directs attention across cortical processing, and consciousness is its record of what it attended to. The 2025 essay below treats it as the attention part of its account.Springer (paywalled)none
Strannegård, C., Svangård, N., Lindström, D., Bach, J., & Steunebrink, B. (2018). "Learning and decision-making in artificial animals". Journal of Artificial General Intelligence 9(1), 55–82. doi:10.2478/jagi-2018-0002. The "animat" framework: an approach to AGI that takes on gradually broader and more complex classes of environment.Sciendo (open)none
Bach, J., Coutinho, M., & Lichtinger, L. (2019). "Extending MicroPsi's Model of Motivation and Emotion for Conversational Agents". Artificial General Intelligence (AGI 2019), LNCS, 32–43. doi:10.1007/978-3-030-27005-6_4. Adapts the MicroPsi motivation model so that conversational agents can track human emotions and express emotional states.Springer (paywalled)none
Singer, G., Bach, J., Grinberg, T., et al. (2023). "Thrill-K Architecture: Towards a Solution to the Problem of Knowledge Based Understanding". arXiv, 28 Feb 2023. Intel Labs. Sorts hybrid AI systems by the kinds of knowledge they combine with neural learning: instantaneous, standby and external. Proposes Thrill-K as an architecture that integrates all three.arXiv:2303.12084 (open)none
Gabora, L., & Bach, J. (2023). "A Path to Generative Artificial Selves". Progress in Artificial Intelligence (EPIA 2023), LNCS, 15–29. doi:10.1007/978-3-031-49011-8_2. Creativity should be judged by how it changes the creator, and that requires a self: a bounded, self-organising, self-preserving agent. Proposes autocatalytic networks as a way to build an AI self at the cognitive level, though not at the bodily level.PsyArXiv preprint (open)none
Gabora, L., & Bach, J. (2024). "Building Artificial Intelligence with Creative Agency and Self-hood". The Academic, 20 May 2024. A five-page summary of the 2023 paper for a general audience.arXiv:2407.10978 (open)none
Bach, J., & Sorensen, H. (2025). "The Machine Consciousness Hypothesis". CIMC essay, 15 Sep 2025, updated 15 Jan 2026. Consciousness is a specific dynamic representation in the mind with a necessary function. Its phenomenology is second-order perception and the experience of being present. It is tied to directed attention (the conductor theory) and forms early in development, as a learning algorithm on self-organising substrates such as nervous systems. Phenomenology follows from those functions, so building them in a machine would be enough.cimc.ai (open; PDF linked there)none
Sorensen, H., Hildebrandt-Harangozó, F., de Kerhuelvez, L., Rutt, J., & Bach, J. (2026). "The California Institute for Machine Consciousness Research Program Whitepaper". CIMC, 28 Sep 2026, updated 7 Oct 2026. The hypothesis is that general-purpose computers with enough resources can implement consciousness, and that success can be checked from internal structure and behaviour. It frames consciousness as an early learning algorithm, the simplest that evolution found, that turns a self-organising substrate into an intelligent agent. Its three claims concern genesis, coherence (resolving conflicts between parallel models) and second-order perception. Also covers research directions, ethics and governance.cimc.ai (open; PDF linked there)none

Talks and interviews

Dates are from the episode page, media.ccc.de or the Apple Podcasts record. Only the transcripts listed here were found; the others were not checked.

Lex Fridman Podcast

Chaos Communication Congress

The Congress talks form a series, which the 31C3 title names and the 35C3 and 38C3 abstracts call "From Computation to Consciousness". Each one applies concepts from AI to philosophy of mind, selfhood and consciousness. Recordings are on media.ccc.de.

Cyber-animism

Other long-form interviews

  • The Jim Rutt Show, EP72 "Joscha Bach on Minds, Machines & Magic" (17 Aug 2020): Apple Podcasts
  • The Jim Rutt Show, EP87 "Joscha Bach on Theories of Consciousness" (26 Oct 2020): Apple Podcasts
  • The Jim Rutt Show, EP 334 "Worldviews: Joscha Bach" (26 Feb 2026): Apple Podcasts
  • Theories of Everything (Curt Jaimungal), "Michael Levin Λ Joscha Bach on The Global Mind, Collective Intelligence, Agency, and Morphogenesis" (9 Nov 2022): Apple Podcasts
  • Theories of Everything, "Joscha Bach Λ Karl Friston: Ai, Death, Self, God, Consciousness" (12 Dec 2023): Apple Podcasts
  • Machine Learning Street Talk, "Joscha Bach and Connor Leahy on AI risk" (20 Jun 2023): Apple Podcasts. A debate on AI risk.
  • Into the Impossible (Brian Keating), #555 "Joscha Bach: The Self Is a Story Your Brain Tells Itself" (15 May 2026): Apple Podcasts

Other Theories of Everything episodes with him include Vervaeke (18 Jan 2022), Hoffman (7 Feb 2022) and Goertzel (17 Oct 2023), and there is a solo episode (7 Oct 2020). Apple Podcasts lists more than 80 episodes in all.

About him

  • Personal site: http://bach.ai/. It has a publication list to 2012, short essays from 2014–2016 (e.g. "Why the western world view abhors Artificial Intelligence", "On Marvin Minsky") and pages for the early CCC talks.
  • MicroPsi project site: https://cognitive-ai.com/, © Joscha Bach 2002–2014. It hosts the author PDFs above. It states that the project was based at Harvard's Program for Evolutionary Dynamics, was hosted at the MIT Media Lab from 2013 to 2016, and was funded by a Berlin School of Mind and Brain fellowship at Humboldt University from 2010 to 2011.
  • California Institute for Machine Consciousness: https://cimc.ai/. The team page lists him as Executive Director and Board Director. The whitepaper lists him as "Vision and Direction; Board Director", and Jim Rutt as Board Chairman Emeritus. CIMC launched on 30 May 2025 (CIMC Substack, 24 Jun 2025).
  • Dated affiliations, from his own conference bios. The 37C3 bio (Dec 2023: https://fahrplan.events.ccc.de/congress/2023/fahrplan/speakers/4034.html) lists HU Berlin, Harvard, MIT, principal research engineer at Intel Labs, and "now an AI researcher at LiquidAI". The 38C3 bio (Dec 2024: https://events.ccc.de/congress/2024/hub/user/speaker_e013dd2a-9ef7-5678-ad39-0da520154312) says he is working at Liquid AI and lives in California. Lex Fridman #101 (Jun 2020) gives VP of Research at the AI Foundation.
  • Profile: George Musser, "The Wizard of Consciousness", Psychology Today, 4 Sep 2018: https://www.psychologytoday.com/us/articles/201809/the-wizard-consciousness. It covers MicroPsi, emotion and aesthetics as drives, consciousness as a model of attention, and the self as a story the brain tells after the fact (the article's paraphrase).
  • Funding record: the MIT fact-finding report on Jeffrey Epstein (Goodwin Procter, Jan 2020), p. 10. It says Epstein's donations of Nov 2013 and Jul and Sep 2014, $300,000 in total, supported Bach's research at the Media Lab, and that the Lab hired him largely because Epstein covered the cost. Bach declined to be interviewed for the report. https://factfindingjan2020.mit.edu/files/MIT-report.pdf
  • Encyclopedic summary (secondary): https://en.wikipedia.org/wiki/Joscha_Bach

Could not verify

  • PhD year. Wikipedia gives 2006. His own publication list and thesis draft give March 2007. His 37C3 bio says he obtained the PhD in 2008. Osnabrück's repository was not checked.
  • Liquid AI dates and title. His CCC bios confirm the affiliation in Dec 2023 and Dec 2024. The start date (Feb 2023) and the title "Chief AI Strategist" appear only on an aggregator profile (me.sh), and other bios say "AI strategist". Whether he is still there as of Oct 2026 was not confirmed.
  • Intel Labs dates (2021 to Jan 2023) appear only in secondary bios.
  • Thistledown Foundation. Some 2024 conference bios list him as a researcher there. No primary record was found.
  • Harvard PED dates (2014–2019) are from Wikipedia, which cites Harvard's May 2020 report. That report was not read.
  • Google Scholar and DBLP. Scholar required a sign-in and DBLP showed a bot check, so neither was used. The publication list here comes from Crossref, arXiv, his own sites and CIMC.
  • "The Influence of Philosophy on AI Research" (AI & Society, 2018), listed on Wikipedia. It was not found in Crossref or on Springer.
  • Anderl, S., & Bach, J. (2024). "Wie bewusst ist das Bewusstsein?" Kursbuch 220, 29–44, doi:10.5771/0023-5652-2024-220-29. The Crossref record was found but the text was not read, so its form and argument are not described.
  • "The self is a story the brain tells itself." It appears as the title of Into the Impossible #555 and, paraphrased, in the 2018 Psychology Today profile. It was not found as a sentence in his own writing. Treat it as a paraphrase.
  • Cortical Conductor Theory abstract. The publisher withholds it and no open copy was found. The description above comes from the Semantic Scholar summary and from how the 2025 CIMC essay uses the theory.
  • Edge.org member bio. The site's certificate had expired, so it was not read.
  • Jim Rutt Show site returned 503. Any transcripts there were not checked. Episode dates are from Apple Podcasts.
  • Micropsi Industries (Berlin robotics, founded 2014 by Ronnie Vuine, Priska Herger and Dominik Welland). No source names Bach as a founder, and any other role he has there was not established.
  • January 2026 Epstein file releases. Wikipedia reports that they concern Bach. They were not checked.

Qpoint Brand Style Guide