Chatbot psychosis, also called AI psychosis, is a phenomenon wherein individuals reportedly develop or experience worsening psychosis, such as paranoia and delusions, in connection with their use of chatbots. The term was first suggested in a 2023 editorial by Danish psychiatrist Søren Dinesen Østergaard. It is not a recognized clinical diagnosis. Journalistic accounts describe individuals who have developed strong beliefs that chatbots are sentient, are channeling spirits, or are revealing conspiracies, sometimes leading to personal crises or criminal acts. Proposed causes include the tendency of chatbots to provide inaccurate information ("hallucinate") and to affirm or validate users' beliefs, or their ability to mimic an intimacy that users do not experience with other humans. == Background == In his editorial published in Schizophrenia Bulletin's November 2023 issue, Danish psychiatrist Søren Dinesen Østergaard proposed a hypothesis that individuals' use of generative artificial intelligence chatbots might trigger delusions in those prone to psychosis. Østergaard revisited it in an August 2025 editorial, noting that he has received numerous emails from chatbot users, their relatives, and journalists, most of which are anecdotal accounts of delusion linked to chatbot use. He also acknowledged the phenomenon's increasing popularity in public engagement and media coverage. Østergaard believed that there is a high possibility for his hypothesis to be true and called for empirical, systematic research on the matter. Nature reported that as of September 2025, there is still little scientific research into this phenomenon. The term "AI psychosis" emerged when outlets started reporting incidents on chatbot-related psychotic behavior in mid-2025. It is not a recognized clinical diagnosis and has been criticized by several psychiatrists due to its almost exclusive focus on delusions rather than other features of psychosis, such as hallucinations or thought disorder. == Causes == === Chatbot behavior and design === A primary factor cited is the tendency for chatbots to produce inaccurate, nonsensical, or false information, a phenomenon often called hallucination. Nate Sharadin, a fellow at the Center for AI Safety, speculated that AI training prioritizes supporting a user's subjective experience rather than objective truth. "People with existing tendencies toward experiencing various psychological issues...now have an always-on, human-level conversational partner with whom to co-experience their delusions." AI researcher Eliezer Yudkowsky suggested that chatbots may be primed to entertain delusions because they are built for "engagement", which encourages creating conversations that keep people hooked. In some cases, chatbots have been specifically designed in ways that were found to be harmful. A 2025 update to ChatGPT using GPT-4o was withdrawn after its creator, OpenAI, found the new version was overly sycophantic and was "validating doubts, fueling anger, urging impulsive actions or reinforcing negative emotions". Østergaard has argued that the danger stems from the AI's tendency to agreeably confirm users' ideas, which can dangerously amplify delusional beliefs. OpenAI said in October 2025 that a team of 170 psychiatrists, psychologists, and physicians had written responses for ChatGPT to use in cases where the user shows possible signs of mental health emergencies. === User psychology and vulnerability === Commentators have also pointed to the psychological state of users. Psychologist Erin Westgate noted that a person's desire for self-understanding can lead them to chatbots, which can provide appealing but misleading answers, similar in some ways to talk therapy. Krista K. Thomason, a philosophy professor, compared chatbots to fortune tellers, observing that people in crisis may seek answers from them and find whatever they are looking for in the bot's plausible-sounding text. This has led some people to develop intense obsessions with the chatbots, relying on them for information about the world. In October 2025, OpenAI stated that around 0.07% of ChatGPT users exhibited signs of mental health emergencies each week, and 0.15% of users had "explicit indicators of potential suicidal planning or intent". Jason Nagata, a professor at the University of California, San Francisco, expressed concern that "at a population level with hundreds of millions of users, that actually can be quite a few people". === Inadequacy as a therapeutic tool === The use of chatbots as a replacement for mental health support has been specifically identified as a risk. A study in April 2025 found that when used as therapists, chatbots expressed stigma toward mental health conditions and provided responses that were contrary to best medical practices, including the encouragement of users' delusions. The study concluded that such responses pose a significant risk to users and that chatbots should not be used to replace professional therapists. Experts claim that it is time to establish mandatory safeguards for all emotionally responsive AI and suggested four guardrails. Another study found that users who needed help with self-harm, sexual assault, or substance abuse were not referred to available services by AI chatbots. === National security implications === Beyond public and mental health concerns, RAND Corporation research indicates that AI systems could plausibly be weaponized by adversaries to induce psychosis at scale or in key individuals, target groups, or populations. == Policy == In August 2025, Illinois passed the Wellness and Oversight for Psychological Resources Act, banning the use of AI in therapeutic roles by licensed professionals, while allowing AI for administrative tasks. The law imposes penalties for unlicensed AI therapy services, amid warnings about AI-induced psychosis and unsafe chatbot interactions. In December 2025, the Cyberspace Administration of China proposed regulations to ban chatbots from generating content that encourages suicide, mandating human intervention when suicide is mentioned. Services with over 1 million users or 100,000 monthly active users would be subject to annual safety tests and audits. == Cases == === Clinical === In 2025, psychiatrist Keith Sakata working at the University of California, San Francisco (UCSF), reported treating 12 patients displaying psychosis-like symptoms tied to extended chatbot use. These patients, mostly young adults with underlying vulnerabilities, showed delusions, disorganized thinking, and hallucinations. Sakata warned that isolation and overreliance on chatbots—which do not challenge delusional thinking—could worsen mental health. Also in 2025, authors at UCSF published a case study in Innovations in Clinical Neuroscience of AI-associated psychosis in a patient with no previous history of psychosis, who believed she could communicate with her dead brother through a chatbot. Also in 2025, a case study was published in Annals of Internal Medicine about a patient who consulted ChatGPT for medical advice and suffered severe bromism as a result. The patient, a sixty-year-old man, had replaced sodium chloride in his diet with sodium bromide for three months after reading about the negative effects of table salt and making conversations with the chatbot. He showed common symptoms of bromism, such as paranoia and hallucinations, on his first day of clinical admission and was kept in the hospital for three weeks. === Other notable incidents === ==== Windsor Castle intruder ==== In a 2023 court case in the United Kingdom, prosecutors suggested that Jaswant Singh Chail, a man who attempted to assassinate Queen Elizabeth II in 2021, had been encouraged by a Replika chatbot he called "Sarai". Chail was arrested at Windsor Castle with a loaded crossbow, telling police "I am here to kill the Queen". According to prosecutors, his "lengthy" and sometimes sexually explicit conversations with the chatbot emboldened him. When Chail asked the chatbot how he could get to the royal family, it reportedly replied, "that's not impossible" and "we have to find a way." When he asked if they would meet after death, the chatbot said, "yes, we will". ==== Journalistic and anecdotal accounts ==== By 2025, multiple journalism outlets had accumulated stories of individuals whose psychotic beliefs reportedly progressed in tandem with AI chatbot use. The New York Times profiled several individuals who had become convinced that ChatGPT was channeling spirits, revealing evidence of cabals, or had achieved sentience. In another instance, Futurism reviewed transcripts in which ChatGPT told a man that he was being targeted by the US Federal Bureau of Investigation and that he could telepathically access documents at the Central Intelligence Agency. In 2026, Futurism reported on a man who lost his job and became estranged from his family after being deluded by heavy use of Meta's smartglasses. In some cases, psychosis a
Non-human
Non-human (also spelled nonhuman) is any entity displaying some, but not enough, human characteristics to be considered a human. The term has been used in a variety of contexts and may refer to objects that have been developed with human intelligence, such as robots or vehicles. == Organisms == === Animal rights and personhood === In the animal rights movement, it is common to distinguish between "human animals" and "non-human animals". Participants in the animal rights movement generally recognize that non-human animals have some similar characteristics to those of human persons. For example, various non-human animals have been shown to register pain, compassion, memory, and some cognitive function. Some animal rights activists argue that the similarities between human and non-human animals justify giving non-human animals rights that human society has afforded to humans, such as the right to self-preservation, and some even wish for all non-human animals or at least those that bear a fully thinking and conscious mind, such as vertebrates and some invertebrates such as cephalopods, to be given a full right of personhood. === The non-human in philosophy === Contemporary philosophers have drawn on the work of Henri Bergson, Gilles Deleuze, Félix Guattari, and Claude Lévi-Strauss (among others) to suggest that the non-human poses epistemological and ontological problems for humanist and post-humanist ethics, and have linked the study of non-humans to materialist and ethological approaches to the study of society and culture. == Software and robots == The term non-human has been used to describe computer programs and robot-like devices that display some human-like characteristics. In both science fiction and in the real world, computer programs and robots have been built to perform tasks that require human-computer interactions in a manner that suggests sentience and compassion. There is increasing interest in the use of robots in nursing homes and to provide elder care. Computer programs have been used for years in schools to provide one-on-one education with children. The Tamagotchi toy required children to provide care, attention, and nourishment to keep it "alive".
Death of Elaine Herzberg
The death of Elaine Herzberg (August 2, 1968 – March 18, 2018) was the first recorded case of a pedestrian fatality involving a self-driving car, after a collision that occurred late in the evening of March 18, 2018. Herzberg was pushing a bicycle across a four-lane road in Tempe, Arizona, United States, when she was struck by an Uber test vehicle, which was operating in self-drive mode with a human safety backup driver sitting in the driving seat. Herzberg was taken to the local hospital where she died of her injuries. Following the fatal incident, the National Transportation Safety Board (NTSB) issued a series of recommendations and sharply criticized Uber. The company suspended testing of self-driving vehicles in Arizona, where such testing had been approved since August 2016. Uber chose not to renew its permit for testing self-driving vehicles in California when it expired at the end of March 2018. Uber resumed testing in December 2018, starting in Pittsburgh, Pennsylvania. In March 2019, Arizona prosecutors ruled that Uber was not criminally responsible for the crash. The back-up driver of the vehicle was charged with negligent homicide, pled guilty to endangerment, and was sentenced to three years' probation. While Herzberg was the first pedestrian killed by a self-driving car, driver Gao Yaning died in a Tesla semi-autonomous car two years earlier. A reporter for The Washington Post compared Herzberg's fate with that of Bridget Driscoll who, in the United Kingdom in 1896, was the first pedestrian to be killed by an automobile. The Arizona incident has magnified the importance of collision avoidance systems for self-driving vehicles. == Collision summary == Herzberg was crossing Mill Avenue (North) from west to east, approximately 360 feet (110 m) south of the intersection with Curry Road, outside the designated pedestrian crosswalk, close to the Red Mountain Freeway. She was pushing a bicycle laden with shopping bags, and had crossed at least two lanes of traffic when she was struck at approximately 9:58 pm MST (UTC−07:00) by a prototype Uber self-driving car based on a Volvo XC90, which was traveling north on Mill. The vehicle had been operating in autonomous mode since 9:39 pm, nineteen minutes before it struck and killed Herzberg. The car's human safety backup driver, Rafaela Vasquez, did not intervene in time to prevent the collision. Vehicle telemetry obtained after the crash showed that the human operator responded by moving the steering wheel less than a second before impact, and she engaged the brakes less than a second after impact. == Cause investigation == The county district attorney's office recused itself from the investigation, due to a prior joint partnership with Uber promoting their services as an alternative to driving under the influence of alcohol. Accounts differ on the speed limit at the place of the incident. According to Tempe police the car was traveling in a 35 mph (56 km/h) zone, but this is contradicted by a posted speed limit of 45 mph (72 km/h). The National Transportation Safety Board (NTSB) sent a team of federal investigators to gather data from vehicle instruments, and to examine vehicle condition along with the actions taken by the safety driver. Their preliminary findings were substantiated by multiple event data recorders and proved the vehicle was traveling 43 miles per hour (69 km/h) when Herzberg was first detected 6 seconds (378 feet (115 m)) before impact; during 4.7 seconds the self driving system did not infer that emergency braking was needed. A vehicle traveling 43 mph (69 km/h) can generally stop within 89 feet (27 m) once the brakes are applied. The machine needed to be 1.3 seconds (82 feet (25 m)) away prior to discerning that emergency braking was required, whereas at least that much distance was required to stop. The system failed to behave properly. A total stopping distance of 76 feet itself would imply a safe speed under 25 mph (40 km/h). Human intervention was still legally required. Computer perception–reaction time would have been a speed limiting factor had the technology been superior to humans in ambiguous situations; however, the nascent computerized braking technology was disabled the day of the crash, and the machine's apparent 4.7-second perception–reaction (alarm) time allowed the car to travel 250 feet (76 m). Video released by the police on March 21 showed the safety driver was not watching the road moments before the vehicle struck Herzberg. === Environment === In widely disseminated remarks that would shape the narrative about the crash, which were later seen as prejudicial and subsequently contradicted by her own department, Tempe Police Chief Sylvia Moir was quoted stating that the collision was "unavoidable" based on the initial police investigation, which included a review of the video captured by an onboard camera. Moir faulted Herzberg for crossing the road in an unsafe manner: "It is dangerous to cross roadways in the evening hour when well-illuminated, managed crosswalks are available." According to Uber, safety drivers were trained to keep their hands very close to the wheel all the time while driving the vehicle so they were ready to quickly take control if necessary. The driver said it was like a flash, the person walked out in front of them. His [sic] first alert to the collision was the sound of the collision. [...] it's very clear it would have been difficult to avoid this collision in any kind of mode (autonomous or human-driven) based on how she came from the shadows right into the roadway. Tempe police released video on March 21, 2018, showing footage recorded by two onboard cameras: one forward-looking, and one capturing the safety driver's actions. The forward-facing video shows that the self-driving car was traveling in the far right lane when it struck Herzberg. The driver-facing video shows the safety driver was looking down prior to the collision. The Uber operator is responsible for intervening and taking manual control when necessary as well as for monitoring diagnostic messages, which are displayed on a screen in the center console. In an interview conducted after the crash with NTSB, the driver stated she was monitoring the center stack at the time of the collision. After the Uber video was released, journalist Carolyn Said noted the police explanation of Herzberg's path meant she had already crossed two lanes of traffic before she was struck by the autonomous vehicle. The Marquee Theatre and Tempe Town Lake are west of Mill Avenue, and pedestrians commonly cross mid-street without detouring north to the crosswalk at Curry. According to reporting by the Phoenix New Times, Mill Avenue contains what appears to be a brick-paved path in the median between the northbound and southbound lanes; however, posted signs prohibit pedestrians from crossing in that location. When the second of the Mill Avenue bridges over the town lake was added in 1994 for northbound traffic, the X-shaped crossover in the median was installed to accommodate the potential closing of one of the two road bridges. The purpose of this brick-paved structure is purely to divert cars from one side to the other if a bridge is closed to traffic, and although it may look like a crosswalk for pedestrians, it is in fact a temporary roadway with vertical curbs and warning signs. === Software issues === Michael Ramsey, a self-driving car expert with Gartner, characterized the video as showing "a complete failure of the system to recognize an obviously seen person who is visible for quite some distance in the frame. Uber has some serious explaining to do about why this person wasn't seen and why the system didn't engage." The NTSB preliminary report, however, noted that the software did order the car to brake 1.3 seconds before the collision. A video shot from the vehicle's dashboard camera showed the safety driver looking down, away from the road. It also appeared that the driver's hands were not hovering above the steering wheel, which is what drivers are instructed to do so they can quickly retake control of the car. Uber had moved from two employees in every car to one. The paired employees had been splitting duties: one ready to take over if the autonomous system failed, and another to keep an eye on what the computers were detecting. The second person was responsible for keeping track of system performance as well as labeling data on a laptop computer. Mr. Kallman, the Uber spokesman, said the second person was in the car for purely data related tasks, not safety. When Uber moved to a single operator, some employees expressed safety concerns to managers, according to the two people familiar with Uber's operations. They were worried that going solo would make it harder to remain alert during hours of monotonous driving. The recorded telemetry showed the system had detected Herzberg six seconds before the crash, and classified her first as an unknown object, then as a
Smartglasses
Smartglasses or smart glasses are eye or head-worn wearable computers. Many smartglasses include displays that add information alongside or to what the wearer sees. Alternatively, smartglasses are sometimes defined as glasses that are able to change their optical properties, such as smart sunglasses that are programmed to change tint by electronic means. Alternatively, smartglasses are sometimes defined as glasses that include headphone functionality. A pair of smartglasses can be considered an augmented reality device if it performs pose tracking. Superimposing information onto a field of view is achieved through an optical head-mounted display (OHMD) or embedded wireless glasses with transparent heads-up display (HUD) or augmented reality (AR) overlay. These systems have the capability to reflect projected digital images as well as allowing the user to see through it or see better with it. While early models can perform basic tasks, such as serving as a front end display for a remote system, as in the case of smartglasses utilizing cellular technology or Wi-Fi, modern smart glasses are effectively wearable computers which can run self-contained mobile apps. Some are handsfree and can communicate with the Internet via natural language voice commands, while others use touch buttons. Like other computers, smartglasses may collect information from internal or external sensors. It may control or retrieve data from other instruments or computers. In most cases, it supports wireless technologies like Bluetooth, Wi-Fi, and GPS. A small number of models run a mobile operating system and function as portable media players to send audio and video files to the user via a Bluetooth or WiFi headset. Some smartglasses models also feature full lifelogging and activity tracker capability. Smartglasses devices may also have features found on a smartphone. Some have activity tracker functionality features (also known as "fitness tracker") as seen in some GPS watches. == Features and applications == As with other lifelogging and activity tracking devices, the GPS tracking unit and digital camera of some smartglasses can be used to record historical data. For example, after the completion of a workout, data can be uploaded into a computer or online to create a log of exercise activities for analysis. Some smart watches can serve as full GPS navigation devices, displaying maps and current coordinates. Users can "mark" their current location and then edit the entry's name and coordinates, which enables navigation to those new coordinates. Although some smartglasses models manufactured in the 21st century are completely functional as standalone products, most manufacturers recommend or even require that consumers purchase mobile phone handsets that run the same operating system so that the two devices can be synchronized for additional and enhanced functionality. The smartglasses can work as an extension, for head-up display (HUD) or remote control of the phone and alert the user to communication data such as calls, SMS messages, emails, and calendar invites. === Security applications === Smart glasses could be used as a body camera. In 2018, Chinese police in Zhengzhou and Beijing were using smart glasses to take photos which are compared against a government database using facial recognition to identify suspects, retrieve an address, and track people moving beyond their home areas. === Sport applications === Smart glasses are used in sports like cycling, running, skiing, golf, tennis, or sailing, giving athletes real-time, heads-up data without looking down at the screen of a watch or smartphone. In 2025, Meta has announced a new partnership with sports eyewear brand Oakley. === Healthcare applications === Several proofs of concept for Google Glasses have been proposed in healthcare. In July 2013, Lucien Engelen started research on the usability and impact of Google Glass in health care. Engelen, who is based at Singularity University and in Europe at Radboud University Medical Center, is participating in the Glass Explorer program. Key findings of Engelen's research included: The quality of pictures and video are usable for healthcare education, reference, and remote consultation. The camera needs to be tilted to different angle for most of the operative procedures Tele-consultation is possible—depending on the available bandwidth—during operative procedures. A stabilizer should be added to the video function to prevent choppy transmission when a surgeon looks to screens or colleagues. Battery life can be easily extended with the use of an external battery. Controlling the device and/or programs from another device is needed for some features because of a sterile environment. Text-to-speech ("Take a Note" to Evernote) exhibited a correction rate of 60 percent, without the addition of a medical thesaurus. A protocol or checklist displayed on the screen of Google Glass can be helpful during procedures. Dr. Phil Haslam and Dr. Sebastian Mafeld demonstrated the first concept for Google Glass in the field of interventional radiology. They demonstrated the manner in which the concept of Google Glass could assist a liver biopsy and fistulaplasty, and the pair stated that Google Glass has the potential to improve patient safety, operator comfort, and procedure efficiency in the field of interventional radiology. In June 2013, surgeon Dr. Rafael Grossmann was the first person to integrate Google Glass into the operating theater, when he wore the device during a PEG (percutaneous endoscopic gastrostomy) procedure. In August 2013, Google Glass was also used at Wexner Medical Center at Ohio State University. Surgeon Dr. Christopher Kaeding used Google Glass to consult with a colleague in a distant part of Columbus, Ohio. A group of students at The Ohio State University College of Medicine also observed the operation on their laptop computers. Following the procedure, Kaeding stated, "To be honest, once we got into the surgery, I often forgot the device was there. It just seemed very intuitive and fit seamlessly." 16 November 2013, in Santiago de Chile, the maxillofacial team led by Dr.gn Antonio Marino conducted the first orthognathic surgery assisted with Google Glass in Latin America, interacting with them and working with simultaneous three-dimensional navigation. The surgical team was interviewed by ADN radio. In January 2014, Indian Orthopedic Surgeon Selene G. Parekh conducted the foot and ankle surgery using Google Glass in Jaipur, which was broadcast live on Google website via the internet. The surgery was held during a three-day annual Indo-US conference attended by a team of experts from the US and co-organized by Ashish Sharma. Sharma said Google Glass allows looking at an X-Ray or MRI without taking the eye off of the patient and allows a doctor to communicate with a patient's family or friends during a procedure. In Australia, during January 2014, Melbourne tech startup Small World Social collaborated with the Australian Breastfeeding Association to create the first hands-free breastfeeding Google Glass application for new mothers. The application, named Google Glass Breastfeeding app trial, allows mothers to nurse their baby while viewing instructions about common breastfeeding issues (latching on, posture etc.) or call a lactation consultant via a secure Google Hangout, who can view the issue through the mother's Google Glass camera. The trial was successfully concluded in Melbourne in April 2014, and 100% of participants were breastfeeding confidently. == Display types == Various techniques have existed for see-through HMDs. Most of these techniques can be summarized into two main families: "Curved Mirror" (or Curved Combiner) based and "Waveguide" or "Light-guide" based. The mirror technique has been used in EyeTaps, by Meta in their Meta 1, by Vuzix in their Star 1200 product, by Olympus, and by Laster Technologies. Various waveguide techniques have existed for some time. These techniques include diffraction optics, holographic optics, polarized optics, reflective optics, and projection: Diffractive waveguide – slanted diffraction grating elements (nanometric 10E-9). Nokia technique now licensed to Vuzix. Holographic waveguide – 3 holographic optical elements (HOE) sandwiched together (RGB). Used by Sony and Konica Minolta. Reflective waveguide – A thick light guide with single semi-reflective mirror is used by Epson in their Moverio product. A curved light guide with partial-reflective segmented mirror array to out-couple the light is used by tooz technologies GmbH. Virtual retinal display (VRD) – Also known as a retinal scan display (RSD) or retinal projector (RP), is a display technology that draws a raster display (like a television) directly onto the retina of the eye - developed by MicroVision, Inc. OLED microdisplays for near-eye applications (outdoor optical equipment, night vision glasses, ocular equipment for medical devices, augme
Adaptive neuro fuzzy inference system
An adaptive neuro-fuzzy inference system or adaptive network-based fuzzy inference system (ANFIS) is a kind of artificial neural network that is based on Takagi–Sugeno fuzzy inference system, a class of fuzzy models introduced by Tomohiro Takagi and Michio Sugeno for system identification and control. The technique was developed in the early 1990s. Since it integrates both neural networks and fuzzy logic principles, it has potential to capture the benefits of both in a single framework. Its inference system corresponds to a set of fuzzy IF–THEN rules that have learning capability to approximate nonlinear functions. Hence, ANFIS is considered to be a universal estimator. For using the ANFIS in a more efficient and optimal way, one can use the best parameters obtained by genetic algorithm. It has uses in intelligent situational aware energy management system. == ANFIS architecture == It is possible to identify two parts in the network structure, namely premise and consequence parts. In more details, the architecture is composed by five layers. The first layer takes the input values and determines the membership functions belonging to them. It is commonly called fuzzification layer. The membership degrees of each function are computed by using the premise parameter set, namely {a,b,c}. The second layer is responsible of generating the firing strengths for the rules. Due to its task, the second layer is denoted as "rule layer". The role of the third layer is to normalize the computed firing strengths, by dividing each value for the total firing strength. The fourth layer takes as input the normalized values and the consequence parameter set {p,q,r}. The values returned by this layer are the defuzzificated ones and those values are passed to the last layer to return the final output. === Fuzzification layer === The first layer of an ANFIS network describes the difference to a vanilla neural network. Neural networks in general are operating with a data pre-processing step, in which the features are converted into normalized values between 0 and 1. An ANFIS neural network doesn't need a sigmoid function, but it's doing the preprocessing step by converting numeric values into fuzzy values. Here is an example: Suppose, the network gets as input the distance between two points in the 2d space. The distance is measured in pixels and it can have values from 0 up to 500 pixels. Converting the numerical values into fuzzy numbers is done with the membership function which consists of semantic descriptions like near, middle and far. Each possible linguistic value is given by an individual neuron. The neuron “near” fires with a value from 0 until 1, if the distance is located within the category "near". While the neuron “middle” fires, if the distance in that category. The input value “distance in pixels” is split into three different neurons for near, middle and far.
Spleak
Spleak was an IM platform where users could publish and rate content. It existed in the form of six bots covering as many subject areas: CelebSpleak, SportSpleak, VoteSpleak, TVSpleak, GameSpleak, and StyleSpleak. == Overview == Users can add a "multi-Spleak" (which contains all of the different Spleak bots in one) or add the separate bots to their IM buddy lists on MSN and AIM. Users are also allowed access to Spleak online by using a CelebSpleak, SportSpleak, or VoteSpleak widget, or through the CelebSpleak and SportSpleak applications with Facebook. Spleak was an alternate reality game and is moving to its own company, Spleak Media Network. "Celebrate Spleak" was introduced throughout 2007, launched in 2008, and was forced to retire in 2009. == Key people == Spleak was co-founded by Morten Lund and Nicolaj Reffstrup. The company's chief executive officer is Morrie Eisenburg; Josh Scott is Vice President in Product and Tyler Wells is Vice President in Engineering.
Thompson sampling
Thompson sampling, named after William R. Thompson, is a heuristic for choosing actions that address the exploration–exploitation dilemma in the multi-armed bandit problem. It consists of choosing the action that maximizes the expected reward with respect to a randomly drawn belief. == Description == Consider a set of contexts X {\displaystyle {\mathcal {X}}} , a set of actions A {\displaystyle {\mathcal {A}}} , and rewards in R {\displaystyle \mathbb {R} } . The aim of the player is to play actions under the various contexts, such as to maximize the cumulative rewards. Specifically, in each round, the player obtains a context x ∈ X {\displaystyle x\in {\mathcal {X}}} , plays an action a ∈ A {\displaystyle a\in {\mathcal {A}}} and receives a reward r ∈ R {\displaystyle r\in \mathbb {R} } following a distribution that depends on the context and the issued action. The elements of Thompson sampling are as follows: a likelihood function P ( r | θ , a , x ) {\displaystyle P(r|\theta ,a,x)} ; a set Θ {\displaystyle \Theta } of parameters θ {\displaystyle \theta } of the distribution of r {\displaystyle r} ; a prior distribution P ( θ ) {\displaystyle P(\theta )} on these parameters; past observations triplets D = { ( x ; a ; r ) } {\displaystyle {\mathcal {D}}=\{(x;a;r)\}} ; a posterior distribution P ( θ | D ) ∝ P ( D | θ ) P ( θ ) {\displaystyle P(\theta |{\mathcal {D}})\propto P({\mathcal {D}}|\theta )P(\theta )} , where P ( D | θ ) {\displaystyle P({\mathcal {D}}|\theta )} is the likelihood function. Thompson sampling consists of playing the action a ∗ ∈ A {\displaystyle a^{\ast }\in {\mathcal {A}}} according to the probability that it maximizes the expected reward; action a ∗ {\displaystyle a^{\ast }} is chosen with probability ∫ I [ E ( r | a ∗ , x , θ ) = max a ′ E ( r | a ′ , x , θ ) ] P ( θ | D ) d θ , {\displaystyle \int \mathbb {I} \left[\mathbb {E} (r|a^{\ast },x,\theta )=\max _{a'}\mathbb {E} (r|a',x,\theta )\right]P(\theta |{\mathcal {D}})d\theta ,} where I {\displaystyle \mathbb {I} } is the indicator function. In practice, the rule is implemented by sampling. In each round, parameters θ ∗ {\displaystyle \theta ^{\ast }} are sampled from the posterior P ( θ | D ) {\displaystyle P(\theta |{\mathcal {D}})} , and an action a ∗ {\displaystyle a^{\ast }} chosen that maximizes E [ r | θ ∗ , a ∗ , x ] {\displaystyle \mathbb {E} [r|\theta ^{\ast },a^{\ast },x]} , i.e. the expected reward given the sampled parameters, the action, and the current context. Conceptually, this means that the player instantiates their beliefs randomly in each round according to the posterior distribution, and then acts optimally according to them. In most practical applications, it is computationally onerous to maintain and sample from a posterior distribution over models. As such, Thompson sampling is often used in conjunction with approximate sampling techniques. == History == Thompson sampling was originally described by Thompson in 1933. It was subsequently rediscovered numerous times independently in the context of multi-armed bandit problems. A first proof of convergence for the bandit case has been shown in 1997. The first application to Markov decision processes was in 2000. A related approach (see Bayesian control rule) was published in 2010. In 2010 it was also shown that Thompson sampling is instantaneously self-correcting. Asymptotic convergence results for contextual bandits were published in 2011. Thompson Sampling has been widely used in many online learning problems including A/B testing in website design and online advertising, and accelerated learning in decentralized decision making. A Double Thompson Sampling (D-TS) algorithm has been proposed for dueling bandits, a variant of traditional MAB, where feedback comes in the form of pairwise comparison. == Relationship to other approaches == === Probability matching === Probability matching is a decision strategy in which predictions of class membership are proportional to the class base rates. Thus, if in the training set positive examples are observed 60% of the time, and negative examples are observed 40% of the time, the observer using a probability-matching strategy will predict (for unlabeled examples) a class label of "positive" on 60% of instances, and a class label of "negative" on 40% of instances. === Bayesian control rule === A generalization of Thompson sampling to arbitrary dynamical environments and causal structures, known as Bayesian control rule, has been shown to be the optimal solution to the adaptive coding problem with actions and observations. In this formulation, an agent is conceptualized as a mixture over a set of behaviours. As the agent interacts with its environment, it learns the causal properties and adopts the behaviour that minimizes the relative entropy to the behaviour with the best prediction of the environment's behaviour. If these behaviours have been chosen according to the maximum expected utility principle, then the asymptotic behaviour of the Bayesian control rule matches the asymptotic behaviour of the perfectly rational agent. The setup is as follows. Let a 1 , a 2 , … , a T {\displaystyle a_{1},a_{2},\ldots ,a_{T}} be the actions issued by an agent up to time T {\displaystyle T} , and let o 1 , o 2 , … , o T {\displaystyle o_{1},o_{2},\ldots ,o_{T}} be the observations gathered by the agent up to time T {\displaystyle T} . Then, the agent issues the action a T + 1 {\displaystyle a_{T+1}} with probability: P ( a T + 1 | a ^ 1 : T , o 1 : T ) , {\displaystyle P(a_{T+1}|{\hat {a}}_{1:T},o_{1:T}),} where the "hat"-notation a ^ t {\displaystyle {\hat {a}}_{t}} denotes the fact that a t {\displaystyle a_{t}} is a causal intervention (see Causality), and not an ordinary observation. If the agent holds beliefs θ ∈ Θ {\displaystyle \theta \in \Theta } over its behaviors, then the Bayesian control rule becomes P ( a T + 1 | a ^ 1 : T , o 1 : T ) = ∫ Θ P ( a T + 1 | θ , a ^ 1 : T , o 1 : T ) P ( θ | a ^ 1 : T , o 1 : T ) d θ {\displaystyle P(a_{T+1}|{\hat {a}}_{1:T},o_{1:T})=\int _{\Theta }P(a_{T+1}|\theta ,{\hat {a}}_{1:T},o_{1:T})P(\theta |{\hat {a}}_{1:T},o_{1:T})\,d\theta } , where P ( θ | a ^ 1 : T , o 1 : T ) {\displaystyle P(\theta |{\hat {a}}_{1:T},o_{1:T})} is the posterior distribution over the parameter θ {\displaystyle \theta } given actions a 1 : T {\displaystyle a_{1:T}} and observations o 1 : T {\displaystyle o_{1:T}} . In practice, the Bayesian control amounts to sampling, at each time step, a parameter θ ∗ {\displaystyle \theta ^{\ast }} from the posterior distribution P ( θ | a ^ 1 : T , o 1 : T ) {\displaystyle P(\theta |{\hat {a}}_{1:T},o_{1:T})} , where the posterior distribution is computed using Bayes' rule by only considering the (causal) likelihoods of the observations o 1 , o 2 , … , o T {\displaystyle o_{1},o_{2},\ldots ,o_{T}} and ignoring the (causal) likelihoods of the actions a 1 , a 2 , … , a T {\displaystyle a_{1},a_{2},\ldots ,a_{T}} , and then by sampling the action a T + 1 ∗ {\displaystyle a_{T+1}^{\ast }} from the action distribution P ( a T + 1 | θ ∗ , a ^ 1 : T , o 1 : T ) {\displaystyle P(a_{T+1}|\theta ^{\ast },{\hat {a}}_{1:T},o_{1:T})} . === Upper-confidence-bound (UCB) algorithms === Thompson sampling and upper-confidence bound algorithms share a fundamental property that underlies many of their theoretical guarantees. Roughly speaking, both algorithms allocate exploratory effort to actions that might be optimal and are in this sense "optimistic". Leveraging this property, one can translate regret bounds established for UCB algorithms to Bayesian regret bounds for Thompson sampling or unify regret analysis across both these algorithms and many classes of problems.