Integrated Information Theory and Instrumental Transcommunication: A Synergistic Exploration
Introduction
Integrated Information Theory (IIT) is a theoretical framework originally developed to explain and quantify consciousness in physical systems. It posits that a system is conscious to the extent that it generates integrated information (denoted Φ) beyond that of its parts, meaning the whole contains irreducible causal-power that “the sum of its parts” cannot account for (IIT & MRI.pdf). On the other hand, Instrumental Transcommunication (ITC) refers to techniques and devices used to ostensibly communicate with non-physical entities or consciousness (e.g. spirits) via electronic instruments. A well-known subset of ITC is Electronic Voice Phenomena (EVP), where anomalous, voice-like signals are detected in audio recording devices without apparent physical sources. This report will summarize and analyze three key IIT papers – IIT 3.0, IIT 4.0, and a recent implementation of IIT on fMRI data – and then discuss how their concepts relate to ITC/EVP. Subsequently, we propose IIT-inspired experimental designs and device concepts for studying or advancing ITC/EVP. Throughout, we highlight underrepresented ITC experimenters like Keith Clark (noted for sound-shaping approaches) and Sonia Rinaldi (known for cross-modal EVP experiments), examining how IIT principles such as Φ, irreducibility, and system partitioning might be operationalized in their context. The goal is to bridge cutting-edge consciousness theory with the frontier of transcommunication research, outlining innovative methodologies for future investigation.
Integrated Information Theory 3.0: Overview and Key Concepts
IIT 3.0 (Oizumi, Albantakis, Tononi 2014) represents a comprehensive formulation of IIT, building on earlier versions 1.0 and 2.0. IIT 3.0 starts from phenomenology – the intrinsic properties of conscious experience – and translates them into postulates about physical systems (IIT & MRI.pdf). Some core tenets include:
- Intrinsic Existence: Consciousness exists intrinsically for the system itself. Correspondingly, IIT treats information as an intrinsic quantity (cause-effect power within the system) rather than an extrinsic observer-dependent measure like Shannon information (IIT & MRI.pdf). A conscious system must have cause–effect power upon itself, i.e. it can influence its own state and be influenced by its prior state (IIT & MRI.pdf).
- Information (Specificity): A conscious experience is specific – it is the particular way it is and not something else. IIT formalizes this by considering how a system’s current state rules out alternative past and future states. In IIT 3.0, “information is treated as causation,” meaning the information in a state is defined by the constraints it places on the system’s possible causes and effects. The system’s mechanisms in a given state specify a cause repertoire (the distribution of past states that could have led here) and an effect repertoire (future states that could follow) – the overlap of these (cause-effect repertoire) quantifies how much the state is specific and informative in a causal sense.
- Integration (Irreducibility): The elements of a conscious system form an integrated whole that cannot be decomposed without losing information. In practice, Φ measures irreducibility: one computes the loss of information (cause-effect power) when the system is partitioned (conceptually cutting the causal connections between parts) (IIT & MRI.pdf). If the whole has cause-effect power greater than any disjoint union of sub-parts, it has Φ > 0. The magnitude Φ quantifies how much extra information the whole generates beyond the sum of its parts (IIT & MRI.pdf). IIT 3.0 introduced the notion of maximally integrated conceptual information (Φ_max), which is the value of Φ for the minimum (least disruptive) partition (also called the minimum information partition, MIP). In simpler terms, Φ_max is the highest integration level achieved by the system in a given state, found by checking all possible ways to split the system and taking the one that minimizes information – the remainder is irreducible integrated information (IIT & MRI.pdf).
- Exclusion: A conscious experience is a definite set of contents – not a mixture of overlapping experiences. In IIT’s physical postulates, this translates to identifying a single “complex” of elements in a given system state that has the maximum Φ. IIT 3.0 asserts that among overlapping sets of elements, the one with the highest Φ_max (maximally irreducible cause-effect structure) corresponds to the conscious “complex,” and others are excluded from being simultaneously conscious ( System Integrated Information – PMC ). This is based on a principle of maximal existence: only the highest irreducible integration truly exists as a unit of consciousness ( System Integrated Information – PMC ). (If two candidate subsets have equal Φ, neither is definite and they exclude each other ( System Integrated Information – PMC ).)
IIT 3.0 was a significant advance in formal rigor. It provided detailed mathematical procedures for calculating Φ for small, discrete systems (typically modeled as binary nodes with a transition probability matrix). Compared to IIT 2.0, which used simpler information measures (like mutual information between past and present states), IIT 3.0’s framework accounted for the full causal structure of a state by considering all possible causes and effects of that state within the system. This made IIT 3.0 far more computationally intensive, but also more aligned with the intuitive notion that consciousness is about being a single unified cause-effect entity.
In summary, IIT 3.0 posits that a physical system has consciousness if it forms an integrated, irreducible cause-effect structure (quantified by Φ). The theory connects to neuroscience by suggesting, for example, that certain brain networks (like thalamo-cortical systems) have high Φ and correspond to consciousness, whereas others (e.g. the cerebellum, despite its many neurons) have low integration and thus do not contribute to conscious experience (IIT & MRI.pdf). This emphasis on integration resonates with the idea that meaningful or purposeful patterns (as in a conscious thought) involve many parts acting in concert, as opposed to independent or random activity.
IIT 4.0: Recent Developments and Refinements
IIT 4.0 is the latest iteration of the theory (formulated over the past few years, c. 2022–2023) that refines and extends IIT 3.0. It incorporates lessons learned and addresses certain theoretical challenges. Notably, IIT 4.0 provides a more rigorous and “axiomatically accurate” formulation of the postulates, ensuring the mathematics matches the intuitive phenomenological claims (Figure 7 from Integrated information theory (IIT) 4.0: Formulating the …). The fundamental principles (intrinsic existence, information, integration, exclusion) remain, but are defined with greater precision:
- The intrinsic existence (intrinsicality) postulate is formalized by requiring that a system has cause-effect power on itself – if nothing within a system can affect or be affected by the rest, it doesn’t intrinsically exist as a unified entity ( System Integrated Information – PMC ).
- The information postulate in IIT 4.0 explicitly states that a system in a particular state must specify one specific cause-effect state out of many possibilities ( System Integrated Information – PMC ). In practice, the theory calculates the “intrinsic information” of a mechanism or system by how selectively it constrains cause and effect repertoires. Factors like determinism and degeneracy of mechanisms come into play: a state that leads to a single definite outcome (deterministic, non-degenerate) yields higher intrinsic information than one that has many possible outcomes or ambiguous causes ( System Integrated Information – PMC ) ( System Integrated Information – PMC ).
- The integration postulate is sharpened to insist that the cause-effect power of a complex must be unified, i.e. it must stem from the whole set of elements acting together, not separably ( System Integrated Information – PMC ). IIT 4.0 continues to quantify this by evaluating partitions: the system integrated information Φ is calculated by finding how much the whole’s cause-effect power exceeds that of its parts in their best possible split configuration (IIT & MRI.pdf).
- The exclusion postulate is given an algorithmic formulation: one identifies maximally irreducible substrates (sets of elements) by their Φ values and assigns consciousness to the highest-Φ “complexes,” excluding overlapping sets ( System Integrated Information – PMC ) ( System Integrated Information – PMC ). IIT 4.0 provides clearer rules for breaking ties and handling systems at different spatiotemporal grains ( System Integrated Information – PMC ), resolving ambiguities from earlier versions about how to choose the “right” level of analysis for consciousness.
In essence, IIT 4.0 doesn’t overturn the core ideas of IIT 3.0 but makes them more comprehensive and exact. It also emphasizes demonstration of concepts through examples. In a recent overview, small model systems (with a handful of binary elements) were analyzed in depth to illustrate how intrinsic information, integration, and exclusion each contribute to identifying conscious “complexes” ( System Integrated Information – PMC ). For instance, researchers showed that increasing the determinism of causal interactions raises a system’s cause-effect information (addressing the information aspect), while introducing feedback loops that cannot be partitioned increases integrated information (addressing the integration aspect), and ultimately a balance of both yields a maximal Φ for a certain grouping of elements (illustrating exclusion of that grouping as the conscious complex) ( System Integrated Information – PMC ) ( System Integrated Information – PMC ). All computations in IIT 4.0 papers are performed with software like PyPhi (a Python library for IIT) on toy systems ( System Integrated Information – PMC ), since evaluating Φ on anything but small, discrete systems remains extremely demanding.
For our purposes, IIT 4.0’s relevance lies in its even clearer framework for quantifying integrated cause-effect structure. It underscores that to detect a consciousness-like process, one should identify a set of elements that together have much more causal influence as a whole than as independent parts, and that this set is maximal. These theoretical refinements could translate into new ways of thinking about ITC device design – for example, designing circuits intended to maximize Φ, or analyzing data for the signature of a maximally irreducible cause-effect structure as a potential sign of a consciousness-related anomaly.
Implementing IIT in fMRI: Measuring Φ in the Brain
While IIT began as a theoretical construct, recent efforts have tried to apply its measures to real neural data. A significant example is a 2023 study by Nemirovsky et al. which developed an implementation of IIT 3.0 in resting-state fMRI data (IIT & MRI.pdf). This paper (provided as “IIT & MRI” in our sources) is particularly important because it shows how integrated information can be measured in practice and how it correlates with levels of consciousness – knowledge that we can extrapolate to ITC/EVP contexts.
Study Context: The experiment involved 17 healthy subjects scanned with fMRI under two conditions: normal wakefulness and sedation with propofol (an anesthetic) (IIT & MRI.pdf). Propofol was used to transiently lower the subjects’ level of consciousness without other major physiological disruptions. The brain’s resting activity was parsed into known functional networks (Resting State Networks such as the Default Mode, Fronto-Parietal Control, Dorsal Attention, etc.), and each network was reduced to 5 regional time-series for computational tractability (IIT & MRI.pdf).
Measuring Φ_max: The authors implemented IIT 3.0’s Φ_max calculation step by step on these fMRI time series. First, the continuous BOLD signals were discretized/binarized per region (e.g. high vs low activity) at each time point. Then for each small network they constructed a state transition probability matrix (TPM) capturing how the pattern of activity transitions from one moment to the next (IIT & MRI.pdf). Using the PyPhi toolkit, they treated each set of 5 brain regions as a network of 5 binary nodes and computed Φ_max for the network’s state at each time point (IIT & MRI.pdf) (IIT & MRI.pdf). Importantly, Φ_max at a given moment is obtained by evaluating every possible bipartition of the network and finding the one that minimizes cause-effect information; the remaining irreducible information is Φ_max (IIT & MRI.pdf). They found (perhaps not surprisingly) that in these highly interconnected brain networks, the partition that cut the network in half (2 vs 3 regions) was always the “weakest link” and the full network acting as a whole yielded the maximal integration (IIT & MRI.pdf). After computing Φ_max for each time point, they averaged over time to get an integrated information measure per condition (awake vs sedated) for each network (IIT & MRI.pdf) (IIT & MRI.pdf).
Comparative Metrics: To validate and compare, the study also computed several other metrics on the same data: Φ* (phi-star, from IIT 2.0’s “decoding” perspective (IIT & MRI.pdf)), Causal Density (CD) via multivariate Granger causality (IIT & MRI.pdf), and average pairwise correlation (IIT & MRI.pdf). Φ* and CD are established measures thought to index conscious level in prior research, though simpler than Φ_max. The idea was to see which measure best tracks the changes in conscious state induced by anesthesia.
Key Findings: The results showed that Φ_max was sensitive to changes in consciousness level, but in a network-specific way. Under propofol sedation, Φ_max dropped significantly in certain cortical networks associated with consciousness, particularly the Fronto-Parietal Network (FPN) and Dorsal Attention Network (DAN) (IIT & MRI.pdf). These two networks are known for high-level cognitive integration and are often called “higher-order” networks. Notably, the changes in Φ_max in FPN and DAN closely mirrored the expected loss of conscious level in sedation (IIT & MRI.pdf). By contrast, some other networks (e.g. Default Mode Network or sensory networks) did not show a significant Φ_max drop with sedation, aligning with the idea that they might be less directly tied to consciousness level (IIT & MRI.pdf) (IIT & MRI.pdf). Causal Density (CD) also dropped under sedation and identified some overlapping networks (FPN, DAN, and additionally parts of Default Mode) as changing (IIT & MRI.pdf), whereas correlation measures did not distinguish conditions well (they changed globally but not in a way specific to conscious vs unconscious states) (IIT & MRI.pdf) (IIT & MRI.pdf). Φ* (the older IIT 2.0 measure) also behaved differently, showing weaker sensitivity to the time-series temporal structure (IIT & MRI.pdf).
Crucially, Φ_max and CD were the most reliable indicators of conscious state – but with a difference: Φ_max, by incorporating integration, was more selective for the core consciousness-supporting networks (IIT & MRI.pdf) (IIT & MRI.pdf). As the authors note, Φ_max extends beyond mere causality by quantifying integrated causation, making it a “more sophisticated measure” than causal density alone (IIT & MRI.pdf). For example, the cerebellum has plenty of causal activity but little integrated info (as IIT predicts), so a pure causality measure might falsely suggest it contributes to consciousness, whereas Φ correctly stays low (IIT & MRI.pdf). Their data reflected this: some networks with many causal interactions but less integration (like parts of the default mode or a retrosplenial network) showed changes in CD but not in Φ_max (IIT & MRI.pdf).
In sum, this fMRI implementation provided empirical support that integrated information (Φ_max) correlates with levels of awareness, and it demonstrated a practical pipeline for estimating Φ from complex, noisy biological data. For our interests, it also establishes techniques we might borrow: using PyPhi on time-series data, binarizing signals, computing Granger causality (for directed interactions), and comparing entropy or correlation as baselines. It is noteworthy that in a prior study cited by the authors, integrated information (Φ) was able to distinguish meaningful stimuli from noise – e.g. brain Φ dropped when subjects viewed random visual noise versus when they watched an engaging movie clip. This suggests Φ is higher for meaningful, structured inputs and lower for unstructured ones, a point of considerable intrigue when we think about EVPs (meaningful voices emerging from what is essentially random noise). In theory, a genuine EVP might cause an increase in integrated information (structured, cause-effect signal) relative to the baseline noise, analogous to how Φ captured the presence of meaningful sensory input in the brain.
Relating IIT Frameworks to ITC and EVP
The theoretical frameworks of IIT 3.0/4.0 provide a language for talking about integrated, irreducible patterns of information, which can be surprisingly relevant when considering ITC and EVP phenomena. While ITC/EVP research is often seen as fringe or pseudoscientific, approaching it through IIT principles could lend a fresh perspective on what to look for and how to interpret unusual signals. Here, we draw parallels between the two domains:
- Integrated Information (Φ) as a Marker of “Presence”: In IIT, Φ > 0 indicates a set of elements forming a unified entity with causal efficacy on itself (a candidate for consciousness). In an ITC scenario, one might hypothesize that if a disembodied consciousness (say, a spirit or other intelligence) attempts to communicate via electronics, it would need to impart some form of integrated cause-effect structure onto the device or signal. Random noise or isolated clicks would have Φ ≈ 0, as they’re easily reducible to independent parts. However, a coherent voice or image arising within the device – especially one that cannot be attributed to any single source or simple linear process – could manifest as a globally integrated pattern. For example, when an EVP voice is said to form out of background static, it often involves simultaneous changes across many frequencies over time (to form phonemes, words, etc.). This is inherently a more integrated use of the signal than the baseline noise. IIT suggests we could quantify this: treat different frequency bands or different sensors as parts of a system and measure Φ – a clear voice should yield a higher Φ (more cross-frequency coupling and irreducible structure) than the random noise states preceding it. In this way, integrated information could serve as an objective indicator of an anomalous, structured phenomenon within ITC data. If an EVP truly involves an external consciousness impressing itself on the system, IIT would predict a jump in Φ as the system plus influence temporarily form a larger integrated whole.
- Irreducibility and System Partitioning: IIT’s notion of irreducibility resonates with how EVP researchers distinguish an authentic phenomenon from artifact. An EVP that is “irreducible” would be one that cannot be explained by splitting the audio into parts (e.g. it’s not just random radio interference on one channel, or a pareidolia of single frequencies). In practice, ITC experimenters might use multiple channels or devices simultaneously to ensure a signal is not coming from a mundane source. IIT provides a systematic way to do this: one could attempt to partition the system and see if the signal disappears or loses coherence. For instance, use two microphones in different locations recording the same session; a real paranormal voice might manifest jointly (with some coherence) on both, whereas a stray radio or electrical noise might hit one recorder only. In IIT terms, the voice would be an integrated effect across the two inputs. We could compute Φ for the joint two-mic system vs. each mic alone: a significantly higher joint Φ would imply the two inputs share integrated information (possibly the voice spans both), indicating an irreducible common cause. Similarly, partitioning could be temporal: if one cuts the audio into time segments (or shuffles the segments), any meaningful cross-temporal structure (like a word) would be destroyed, reducing Φ. This is analogous to the “temporal control” the fMRI study did (scrambling time points to break causal structure, which caused Φ_max to drop) (IIT & MRI.pdf) (IIT & MRI.pdf). In EVP analysis, one might shuffle or invert phases to test if a perceived voice is an integrated pattern or just isolated bursts. The expectation is that a true integrated phenomenon won’t survive partitioning intact, because its essence lies in the combination of parts. Using irreducibility as a criterion might help filter out false positives (e.g. a blip that only shows up on one frequency band and nowhere else might be considered reducible and likely noise).
- Cause-Effect Structure vs. Simple Causality: One challenge in ITC research is distinguishing meaningful, perhaps intelligent communications from random correlations or mundane causes. IIT’s emphasis on cause-effect power rather than just correlation could be illuminating. A device might register some correlated fluctuations (for example, a spike in an EM meter whenever a particular word is spoken by an investigator), but correlation alone can mislead. IIT would encourage analyzing the causal structure: is there a bidirectional influence? Does the appearance of the anomalous signal demonstrably influence the device’s future states in a way that suggests a feedback loop or memory? In EVP, some advanced experiments use feedback loops (such as voices formed in the output of a device fed back into its input). If a conscious agent were manipulating the system, one might see complex causal loops in the data (for instance, the agent uses the device’s previous output state to shape the next state, etc.). Tools like Granger causality or even full IIT cause-effect analysis can be employed on multi-channel recordings to see if there are directed influences that cannot be explained by chance. In fact, the fMRI study’s use of Causal Density (CD) is analogous – they found that causality measures alone picked up many interactions, some not related to consciousness (IIT & MRI.pdf). By analogy, an EVP setup might show various causal links (say between environmental noise and recorder input), but only a subset that also shows high integration might indicate a conscious influence. In short, IIT’s framework suggests looking for a convergence of high causality and high integration as the sign of genuine ITC phenomena. This could differentiate, for example, a simple radio interference (which might show a cause – a radio transmitter – affecting your recorder, but it’s not integrated with anything else in your system) from a purported spirit voice (which might involve the device’s internal noise, feedback path, etc. all being modulated together as one cause-effect entity).
- Physical Substrates and Complexes: A provocative implication of IIT is that if a conscious mind were to interact with a device, for that moment the mind+device could form one larger system (a temporary “complex” in IIT terms). Traditionally, ITC researchers speak of spirits manipulating electronics; IIT would describe it as two systems coupling into one integrated system with a higher Φ than either alone. This is speculative, but thinking this way could inspire experimenters to create conditions conducive to forming such joint complexes. For instance, providing a medium that is easily influenced and has many degrees of freedom (high potential integration) might be more likely to “host” a conscious influence than a rigid, simple device. This is somewhat analogous to physical mediumship in parapsychology, where a physical medium provides ectoplasm or energy for a spirit to manifest. Here, one might make an electronic “medium” device with rich dynamics and see if it can be driven into integrated patterns by an external mind. Keith Clark’s approach of “sound shaping,” described below, can be seen in this light – by providing sound input that has properties resembling human speech (rich frequency content, formant-like structures), he may be effectively increasing the likelihood of forming an integrated voice pattern when influenced, as opposed to using pure white noise which is harder to mold.
With these conceptual bridges laid out, we now examine some historical and contemporary ITC techniques – particularly those by Keith Clark and Sonia Rinaldi – through the lens of IIT principles.
ITC Techniques: Past and Present
ITC has a fascinating history of inventors and experimenters who developed devices to capture inexplicable communications. Early pioneers like Friedrich Jürgenson and Konstantin Raudive in the 1960s popularized EVP using tape recorders, and later researchers extended ITC to television/video (Klaus Schreiber’s video feedback loops in the 1980s) and radio-based devices (such as Spiricom by George Meek and William O’Neil, which in 1982 purported to facilitate two-way conversations via a set of tone generators). Many of these devices rely on generating some form of stochastic input (noise, audio fragments, visual static) and hoping that an external intelligence can manipulate it into coherent speech or images. Contemporary ITC often uses “spirit boxes” (radios rapidly scanning frequencies) or apps that provide random speech snippets.
Two underrepresented figures in this field, Keith Clark and Sonia Rinaldi, have made notable contributions with innovative methodologies. We highlight their approaches and consider how IIT might interpret or enhance them.
Keith Clark’s Sound Shaping Techniques
Keith J. Clark is an ITC researcher who has focused on the audio aspect of transcommunication. He is known for developing sound shaping or “energy shaping” methods for EVP. The basic idea is to provide pre-designed sound sources that contain rich acoustic energy which could be shaped into speech by an external influence. For example, instead of relying on raw white noise (which is flat and lacks the resonant qualities of human voice), Clark experiments with noise that has human speech formants or fragments of phonemes. By coloring the noise – e.g. filtering it to emphasize speech frequencies or using chopped-up human speech syllables as a feedstock – the hypothesis is that it becomes easier for a spirit to impress a meaningful voice on it (much like how it’s easier to sculpt clay than to sculpt dry sand).
Clark has been involved in ITC initiatives such as the Metascience Foundation and the community Varanormal, aiming to modernize ITC research with digital tools. His sound-shaping devices often use computer software or custom electronics to generate the tailored audio streams. In practice, one of his techniques might involve playing an amorphous human-like vocal mumble (unintelligible but voice-like sounds) in the background while recording. The expectation is that paranormal voices will utilize this “almost voice” energy and clarify it into actual words (a process sometimes reported by EVP enthusiasts). From an IIT perspective, what Clark is doing is increasing the potential for integration in the audio signal – the base signal already has a complex, speech-like structure (hence non-random and easier to integrate), so any additional modulation by an external agent could quickly lead to an irreducible pattern (a spoken word) that stands out. White noise, in contrast, has high entropy and low integration; an EVP within white noise might be harder to distinguish or require more drastic manipulation (hence potentially lower signal-to-noise). By lowering the entropy of the starting material (making it more ordered but still random enough to be flexible), sound shaping might effectively amplify the Φ of any emerging voice. In other words, Clark’s approach is like preparing a canvas that is already textured in the direction of human speech, so that even a small “push” from a purported spirit yields an integrated, intelligible result.
Clark also emphasizes live real-time experiments and collaborative monitoring of ITC sessions. This involves streaming audio in real time and allowing human operators to detect voices. One could imagine coupling this with real-time computation of complexity metrics: for instance, calculating the spectral entropy of the output in real time to see if it dips (indicating more structure) when an EVP is heard, or using algorithms to detect when the audio’s statistical properties deviate from the baseline shaped noise. Indeed, an IIT-inspired metric running alongside could alert an experimenter, “there is an increase in integrated information in the audio right now,” even before the words are fully discerned by ear. Clark’s techniques lay the groundwork for such an augmented approach, as the audio is already in a digital form conducive to analysis.
Sonia Rinaldi’s Cross-Modal Experiments
Sonia Rinaldi, based in Brazil, is a prominent contemporary ITC researcher known for her bold experimental EVP research using unconventional and cross-modal methods. She co-founded IPATI (Instituto de Pesquisas Avançadas em Transcomunicação Instrumental) and has reported pioneering results in both audio and video ITC. Rinaldi often works with modified electronics – for example, repurposing telephone devices, radios, baby monitors, or building custom circuits – and she frequently explores cross-modal correlations, such as capturing audio and video phenomena in tandem.
One of Rinaldi’s hallmark experiments involves capturing purported images of deceased persons on video while simultaneously conducting audio ITC sessions. She has used a video camera aimed at distorted visual patterns (like a screen displaying random noise, or light shining through moving water) and claimed that faces of discarnate people manifest in the video frames. At the same time, she would attempt to record their voices. This is a classic example of cross-modal ITC: using both auditory and visual channels as complementary avenues for communication. From a data perspective, she ends up with two streams (audio and video) that can be analyzed for anomalies. IIT principles nudge us to ask: are these streams independent, or do they form an integrated system during the purported communication? If an unseen consciousness is indeed coming through, one might expect some coupling – e.g. the timing of an audio voice might coincide with the appearance of a face on video, or certain audio features might correlate with video pixel changes. Measuring the integrated information across audio and video could be a fascinating approach. We could treat the audio waveform and the video frame (or some features of it) as parts of a single state vector and attempt to compute a cross-modal Φ. While computing full IIT Φ for high-dimensional video+audio is impractical, even simpler metrics could be used: for instance, mutual information between audio and video signal features at key moments, or multi-modal entropy (entropy of the joint audio-video distribution vs separate). Rinaldi’s work implicitly suggests that meaningful communications might span multiple physical media – which is essentially saying there is a larger cause-effect structure that integrates those media. This aligns with IIT’s notion that a conscious cause will imprint an integrated signature.
Another angle of Rinaldi’s research is using electronic speech synthesis as input. In some experiments, she provides an artificial voice as a starting point – for example, a computer-generated gibberish or foreign-language speech – and then records the output, wherein intelligible phrases (in the target language, often Portuguese, Rinaldi’s language) allegedly emerge. This is somewhat akin to Clark’s sound shaping, but explicitly using a cross-language or transformed base: e.g. using Spanish phonemes as raw material to produce Portuguese words (which were not present originally). Technically, what’s happening is a transformation of information from one form to another, presumably guided by an external mind. If we ignore the paranormal aspect, it’s still an information processing scenario: some process is mapping the input gibberish to coherent output. One could attempt to characterize this process. For instance, one might analyze the cause-effect relationships between the input and output signals. Does the output have features that cannot be directly explained as a linear combination of the input? If yes, that indicates an irreducible addition of information. Indeed, Rinaldi often finds entirely new words or sentences not present in the input audio, which means new information appeared in the system. An information-theoretic analysis could quantify this: measure the information gain from input to output (using, say, conditional entropy or source–output mutual information). In an IIT sense, if the system (input + whatever process + output) has a significant cause-effect structure that isn’t attributable just to the input, that suggests an external cause injecting information (which believers interpret as spirit influence).
Rinaldi’s use of modified devices – from hacked radios to ultrasound transducers – also indicates she is searching for the optimal “vehicle” for communication. Some devices may be more conducive to integrated behavior (just as certain brain regions integrate to support consciousness). For example, using an ultrasound carrier (frequencies above human hearing) that is then modulated might create a kind of high-frequency “canvas” that can later be translated down into audible range, perhaps capturing subtler influences. It’s speculative, but maybe ultrasound offers less noisy channels that only produce audible output when modulated in very specific ways. That specificity could correspond to higher integration when it does occur.
In summary, Rinaldi’s experiments push the boundaries of modality and transformation in ITC, giving us multiple channels of data. They beg for an analytical approach that can handle the complex interdependence of these channels – precisely the sort of approach IIT and related complexity measures are built for.
IIT-Inspired Experimental Designs for ITC/EVP
Bridging IIT with ITC suggests a host of new experiments and device concepts. Here we propose several designs and methodologies, inspired by IIT’s principles, that could be employed to study or enhance instrumental transcommunication and EVP detection. Each proposal aims to operationalize concepts like integrated information, irreducible causality, and complexity in the context of ITC. We remain open-minded about modalities – audio, radio, video, electromagnetic sensors, etc. – as IIT principles are quite general.
1. Multi-Channel Φ Measurement in EVP Sessions
Concept: Set up an EVP recording session with multiple parallel channels of input, and analyze them for integrated information in real time. For example, use two or more microphones in different locations of the room (or one normal mic and one radio-frequency receiver, etc.) recording simultaneously. Most of the time, these channels will pick up independent noise (yielding very low Φ between them). If an anomalous voice or sound occurs that is truly of unknown origin, it might manifest across multiple channels (even if faintly). We can treat the multi-channel audio data as a single system and compute metrics like Φ (for a simplified discretized version) or other complexity measures. A practical approach is to discretize each channel’s signal amplitude to a binary high/low each millisecond, creating a state vector (e.g. [mic1_high, mic2_high, mic3_high]) and use PyPhi to compute integrated information for small time windows of states (IIT & MRI.pdf) (IIT & MRI.pdf). A spike in Φ would indicate a moment where the channels became interdependent in an irreducible way – a strong hint of a cross-channel phenomenon. Even without full Φ, simpler metrics could serve as proxies: e.g. multi-channel entropy (entropy of the joint signals vs sum of individual entropies) or synchrony/causal metrics (like a sudden increase in Granger causality between microphones). This experiment essentially builds an “integrated EVP detector”: rather than just capturing a sound, it evaluates if that sound had a complex, unified imprint across the sensor array. If, say, a voice is only on one mic and not the others, the system’s Φ might remain low (suggesting a local source); but if it registers on all, even faintly, forming a pattern, Φ would be higher, pointing toward a more mysterious integrated source. This approach also guards against false positives by requiring coherence across channels (which random interference is less likely to produce).
2. Complexity-Triggered Voice Extraction
Concept: Use complexity measures to detect and extract EVP signals from noise. This is a software/hardware approach that could augment devices like spirit boxes or noise generators. As audio is recorded, a sliding-window analysis computes measures such as Lempel-Ziv complexity, spectral entropy, or even IIT’s Φ* (phi-star from IIT 2.0, which has practical algorithms (IIT & MRI.pdf) (IIT & MRI.pdf)). When the complexity or integrated information of the audio deviates beyond a threshold – indicating the presence of an unusually structured pattern – the system flags that segment and could apply additional enhancement (like adaptive filtering) to clarify it. The rationale comes from the fMRI result where meaningful stimuli raised integrated information. In EVP, an intelligible voice is the “meaningful stimulus” amid noise. By continuously monitoring complexity, we can create a kind of realtime alarm for potential EVP. This could be implemented in an EVP recording app that shows a live graph of audio entropy and Φ-like metrics; when a drop in entropy (more order) coincides with a rise in Φ or causal interactions, the app could automatically mark that timeframe for the user to review. Such a tool reduces the reliance on subjective listening by giving quantitative hints of where something anomalous may lie. It effectively treats the appearance of integrated information as the signature of a possible communication. Moreover, by logging these metrics, one could analyze patterns across sessions – e.g., do EVP-rich sessions consistently show lower average entropy or higher causality than control sessions? This would bring statistical rigor to EVP research.
3. Feedback Loop Devices with Maximal Integration
Concept: Create an electronic device or circuit that is intentionally designed to have a high Φ and rich cause-effect loops, then use it as a medium for ITC. The hypothesis is that a device which itself approaches a complex, conscious-like state might be more sensitive or responsive to external consciousness influence (almost like meeting the phenomenon halfway). One implementation could be a network of logic gates or neurons on a chip arranged in a recurrent loop topology (e.g. a small field-programmable gate array configured with numerous feedback loops). Using PyPhi or similar, one can design this network to maximize integrated information (for example, prior work suggests certain architectures like fully connected loops yield higher Φ than feedforward ones (IIT 3.0.pdf) (IIT 3.0.pdf)). Once built, this “Φ-circuit” would be set to run with some internal noise or oscillation. If a spirit or conscious intent tries to communicate, perhaps it could more easily modulate a system that is already in a dynamic, integration-ready state. We would monitor the state of this circuit (which could be represented by logic outputs or analog signals) for deviations that carry information. Essentially, this is like building a simple artificial brain that is teetering on the edge of different states (high integration, high sensitivity) and seeing if something can nudge it. The output could be interpreted via attached peripherals: e.g. the circuit’s changing states could be converted into sound (so you “hear” its internal activity) or shown as patterns on a screen. Notably, if any unusual structured pattern emerges (like coherent oscillations corresponding to Morse code or speech elements), it would be extremely intriguing. From an IIT viewpoint, we’d also check if the Φ of the circuit goes even higher during such an event (meaning the external influence actually made the system more integrated – a possible sign of two systems interacting as one). We might call this device a “consciousness-coupled circuit” for ITC. At the very least, regardless of paranormal aspects, studying high-Φ circuits and their noise modulation could yield insights into how to detect subtle signals in complex systems.
4. Cross-Modal Entanglement Experiments
Concept: Building on Rinaldi’s cross-modal approach, set up an experiment where multiple different sensors (audio, video, electromagnetic, etc.) operate simultaneously, and look for moments of entangled behavior among them. For instance, in a controlled dark room, run an audio noise source and a video feedback loop together. Use a microphone to record audio output and a camera for the video output. Apply an algorithm to continuously compute the mutual information between audio and video streams – essentially checking if the audio and video become statistically dependent. Under normal conditions, a sound and an image stream (especially if one is random noise and the other is visual noise) should have near-zero mutual information (they are independent processes). If at some point they show a spike in mutual information or cross-correlation – for example, a specific sound coincides with a specific change in the image – that could be a clue of a single source influencing both (what IIT would describe as a common cause-effect structure spanning both modalities). To make it more specific, one could use modulated signals: perhaps the audio is white noise but the video is being modulated by the audio (or vice versa) in a known way; then an external influence might disturb that modulation in a coherent fashion. Another variant: have two separate random event generators (like two hardware random number generators), one controlling an audio noise volume, another controlling a visual effect. These should be independent – unless something injects information making them momentarily synced. This is inspired by experiments in parapsychology with random number generators (RNGs) where anomalies were reported during intense events or intentions. Here we up the ante by having two RNG-based systems and looking for correlated deviations – essentially an integrated information between two random streams. If consciousness (not embodied in our space) can affect random systems, affecting two in a coordinated way is even more convincing. We’d measure joint entropy vs single entropies to quantify any coupling. If a significant decrease in joint entropy (indicating increased order across the streams) occurs at a time when, say, an EVP voice or some subjective event is noted, that’s noteworthy. This kind of cross-modal, multi-sensor approach treats the whole experimental setup as one big system where an influence might create an integrated fluctuation – something IIT strongly guides us to consider.
5. Causal Signature Analysis of ITC Signals
Concept: Develop a pipeline to analyze recorded ITC/EVP signals for their causal structure using techniques from IIT and computational neuroscience. After an ITC session, especially one with alleged phenomena, we feed the multivariate data (audio channels, maybe environmental sensors too) into a toolkit that performs analyses like: Granger causality network mapping, transfer entropy calculations between channels, and perhaps even attempts a simplified IIT cause-effect analysis (using a binned discrete model). The goal is to produce a “causal signature” of the session – essentially a graph or set of metrics that say how information flowed among components. We then compare this between “active” segments (with voices or anomalies) and “quiet” baseline segments. A hypothesis might be: during an anomalous event, the causal graph will show tighter integration – perhaps all sensors suddenly converge in influence, or a normally dormant pathway (e.g. from an EM spike to the audio recorder) becomes active. If we had enough data, we could compare sessions with claimed ITC contact vs. control sessions with just random noise, to see if their causal signatures differ systematically. This approach, while requiring careful statistical rigor, could lend a more objective characterization to EVP recordings. Instead of solely “I hear a voice,” one could say “the voice segment exhibited a causal density 30% higher than any noise segment” or “the audio and environmental sensor showed a directed coupling that is otherwise absent” – quantitative hallmarks of something unusual. In IIT terms, we’re checking if in those moments the system behaved more like a single integrated system (higher causal density and Φ) versus disjoint parts. Borrowing the fMRI study’s terms, we’d examine if measures like Φ_max and causal density (CD) go up during purported communication. If both go up together, that’s a strong indicator consistent with an integrated causal event (mirroring how in the brain both Φ_max and CD dropped when consciousness was reduced (IIT & MRI.pdf) (IIT & MRI.pdf), here we might see them rise when a “consciousness” possibly enters or influences the system).
Each of these experimental ideas takes inspiration from IIT to either detect ITC phenomena more robustly or to create conditions more amenable to such phenomena. It’s important to note that applying IIT outside of neuroscience is speculative – we must be cautious not to over-interpret noise as Φ. However, the beauty of these proposals is that they are testable and invite objective analysis. Even if one is skeptical of the paranormal, these methods could simply improve signal detection of any weak, structured signal hidden in noise.
Conclusion
By examining Integrated Information Theory 3.0 and 4.0 alongside the esoteric realm of Instrumental Transcommunication, we have drawn a number of intriguing connections. IIT provides a rigorous framework for understanding what it means for information to be meaningfully structured and unified, and the IIT-based fMRI study demonstrates that these measures of integration correlate with the presence of consciousness and meaningful stimuli in the brain (IIT & MRI.pdf). Translating this to ITC/EVP, we infer that if there is any legitimacy to purported spirit communications, they too should manifest as meaningful, structured, and integrated patterns within physical devices – patterns that we can, in principle, detect with the right metrics.
We discussed how IIT’s key principles – integrated information (Φ), irreducibility via system partitioning, and complex cause-effect structures – could be leveraged to design better ITC experiments. In doing so, we also highlighted the innovative work of Keith Clark and Sonia Rinaldi, showing that many of their techniques (sound shaping, cross-modal data capture) intuitively seek to enhance or capture integrated phenomena, even if not originally framed in those terms. By reframing their approaches with IIT, we open up possibilities like measuring cross-channel Φ or identifying when an EVP signal is truly irreducible to simpler explanations.
The experimental designs proposed are admittedly exploratory, straddling the boundary of mainstream science and fringe research. Yet, they are built on solid principles of information theory and complexity science. Even if one set aside the question of spirits, these experiments could yield insights into how highly integrated signals can emerge from noise and how to detect them – knowledge applicable to fields like random signal analysis, cognitive computing, and beyond. On the other hand, if any of these experiments did produce results consistent with a conscious influence (e.g. clear increases in Φ or causal interactions coincident with anomalous phenomena), it would not only be a breakthrough for ITC research but might also inform IIT itself (perhaps expanding the scope of what kinds of systems can exhibit integrated cause-effect structures associated with mind).
In conclusion, Integrated Information Theory offers a compelling quantitative lens to re-examine ITC and EVP. It encourages us to look for integration and causal structure where previously researchers only had ears and eyes. As technology and theory progress, the dream of establishing reliable instrumental communication – if at all possible – may benefit from this convergence of neuroscience-inspired metrics with the imaginative experimentation of ITC pioneers. By being systematic and interdisciplinary in our approach, we move a step closer to discerning signal from noise, whatever its source, and to understanding the nature of that signal in terms of information and integration. Such an understanding is valuable, be it for confirming an extraordinary hypothesis or for finally laying it to rest with scientific clarity.
Sources:
- Oizumi, Masafumi, Larissa Albantakis, and Giulio Tononi. “From the phenomenology to the mechanisms of consciousness: integrated information theory 3.0.” PLoS Comput. Biol. 10(5): e1003588 (2014). (IIT & MRI.pdf)
- Albantakis, Larissa et al. “Integrated Information Theory (IIT) 4.0: formalizing postulates and causality for consciousness.” (Preprint ArXiv:2212.14787, 2022). ( System Integrated Information – PMC )【22†L589-L597# Integrated Information Theory and Instrumental Transcommunication: A Synergistic Exploration