Category: AI

Posted on

AI Sorting

Algorithms do the work behind AI systems. Therefore a basic understanding of how algorithms work is helpful to gauge the potential, risks and performance of such systems. The speed of computers determines the for example the amount of data you can sort at a reasonable time. Efficiency of the algorithm is an other factor. Here we go, we are already a bit absorbed into the the sorting as purely intellectual exercise. The website of Darryl Nester shows a playful programming exercise to sort numbers from 1 to 15 in a fast way (Link to play sorting). If you watch the sorting as it runs you realize that programs are much faster than us in such simple numeric tasks. Now think of applying this sorting routine or algorithm to a process of social sorting. The machine will sort social desirability scores of people’s behavior in the same simple fashion even for thousands of people. Whether proposed AI systems in human interaction or of human resource departments make use of such sorting algorithms we do not know. Sorting applicants is a computational task, but the data input of personal characteristics is derived from another more or less reliable source. Hence, the use of existing and newly available databases will create or eliminate bias. Watching sorting algorithms perform is an important learning experience to be able to critically assess what is likely to happen behind the curtains of AI.

Posted on

AI and dialect

The training of Large Language Models (LLM) uses large data sets to learn about conventions of which words are combined with each other and which ones are less frequently employed in conjunction. Therefore, it does not really come as a surprise that training which uses standardised languages of American English might not be as valid for applications that receive input from minority languages or dialects. The study forthcoming in the field of Computer science and Language by Hofmann et al. (Link) provides evidence of the systematic bias against African American dialects in these models. Dialect prejudice remains a major concern in AI, just like in the day-to-day experiences of many people speaking a dialect. The study highlights that dialect speakers are more likely to be assigned less prestigious jobs if AI is used to sort applicants. Similarly, criminal sentences will harsher for speakers of African American. Even the more frequent attribution of death sentences for dialect speakers was evidenced.
If we translate this evidence to wide-spread applications of AI in the workplace, we realise that there are severe issues to resolve. The European Trade Union Congress (ETUC) has flagged the issue for some time (Link) and made recommendations of how to address these shortcomings. Human control and co-determination by employees are crucial in these applications to the world of work and employment. The need to justify decision-making concerning hiring and firing limit discrimination in the work place. This needs to be preserved in the 21st century collaborating with AI. The language barriers like dialects or multiple official languages in a country ask for a reconsideration of AI to avoid discrimination. Legal systems have to clarify the responsibilities of AI applications before too much harm has been caused.
There are huge potentials of AI as well in the preservation of dialects or interacting in a dialect. The cultural diversity may be preserved more easily, but discriminatory practices have to be eliminated from the basis of these models otherwise they become a severe legal risk for people, companies or public services who apply these large language models without careful scrutiny.
(Image AI BING Designer: 3 robots are in an office. 2 wear suits. 1 wears folklore dress. All speak to each other in a meeting. Cartoon-like style in futuristic setting)

Posted on

AI and S/he

There was hope that artificial intelligence (AI) would be a better version of us. Well, so far that seems to have failed. Let us take gender bias as a pervasive feature even in modern societies, let alone the societies in medieval or industrial age. AI tends to uphold gender biases and might even reinforce them. Why? A recent paper by Kotek, Dockum, Sun (2023) explains the sources for this bias in straightforward terms. AI is based on Large Language Models. These LLMs are trained using big detailed data sets. Through the training on true observed data like detailed data on occupation by gender as observed in the U.S. in 2023, the models tend to have a status quo bias.
This means they abstract from a dynamic evolution of occupations and the potential evolution of gender stereotypes over years. Even deriving growing or decreasing trends of gender dominance in a specific occupation the models have little ground for reasonable or adequate assessment of these trends. Just like thousands of social scientists before them. Projections into the future or assuming a legal obligation of equal representation of gender might still not be in line with human perception of such trends.
Representing women in equal shares among soldiers, 50% of men as secretaries in offices appears rather utopian in 2024, but any share in-between is probably arbitrary and differs widely between countries. Even bigger data sets may account for this in some future day. For the time being these models based on “true” data sets will have a bias towards the status quo, however unsatisfactory this might be.
Now let us just develop on this research finding. Gender bias is only one source of bias among many other forms of bias or discriminatory practices. Ethnicity, age or various abilities complicate the underlying “ground truth” (term used in paper) represented in occupation data sets. The authors identify 4 major shortcoming concerning gender bias in AI based on LLMs: (1) The pronouns s/he were picked even more often than in Bureau of Labor Statistics occupational gender representations; (2) female stereotypes were more amplified than male ones; (3) ambiguity of gender attribution was not flagged as an issue; (4) when found out to be inaccurate LLMs returned “authoritative” responses, which were “often inaccurate”.
These findings have the merit to provide a testing framework for gender bias of AI. Many other biases or potential biases have to be investigated in a similarly rigorous fashion before AI will give us an authoritarian answer, no I am free of any bias in responding to your request. Full stop.

Posted on

Personal Health

Most people would agree, health is a personal issue. From the onset of life, we have package of genes that predetermine a number of factors of our personal health. Epigenetics has taught us there are many factors to take into account additionally. Environmental factors have huge impacts as well. Improvements in the availability of medical devices in the hands of individuals as well as AI systems on portable devices like smartphones facilitate the monitoring of personal health. Several indicators of early-onset of illness can be retrieved from such devices. Dunn et al. (2024) show that prior to the onset of symptoms of Covid-19 or influenza portable devices can indicate the presence of infections through indicators of resting body temperature, heart rate/min, heart rate variability/millisecond or respiratory rate/min. Combined with the indicators of air quality, indoors as well as outdoors, the presence of allergens a much more personalized data set emerges which can easily be part of an AI-assisted diagnosis. More abundant personal health data and analytical power allows remote and digital health applications to inform patients, medical doctors and the public at large. Digital health technologies are only at the beginning to unfold their potential. Prevention becomes more feasible using such devices, medical professionals should be allowed to focus on interpretation of data and treatment rather than simple data gathering. Thinking about digital health technologies points in the direction of dealing with climate and environmental hazards as sickening causes more forcefully. Personal medicine and personal health are, after all, still heavily dependent on health and safety at work, commuting practices and all sorts of pollution. Personal health, however, is a good starting point to raise awareness of the potentials of digital health technologies to better our lives.
(Image: AI MS-Copilot: 2 robots run in a city. They sweat. The air is full of smog. 2 other robots rest near pool. All look at their wrist watch showing heart beats)

Posted on

Error 444

The error message 444 is a not so rare encounter while surfing on the web. The error code 444 stands for the message that from the side of the server the connection is closed without any additional message. The occurrence leaves you without further indication of what exactly went wrong in building a connection to a web service or website. You just simply get shut out and that’s it. It may be tough on you if concerns your online banking or other access to vital services delivered through the internet.
In organization science and social science there are many new studies dealing with the finding, dealing, coping or handling errors. It has become much more acceptable to deal openly with errors, blunders or failures. From an individual as well as organizational perspective the cultures that deal openly with these events seem to have a certain advantage to overcome the consequences of errors at all or faster than others.
In some biographies failures are part of the lessons learned throughout life. It is deemed important to acknowledge failures or paths not taken for better or worse. Organizations just like individuals seem to learn more intensively from their failures or omissions than from everything seemingly running smoothly. Learning curves are faster also for “machine learning” if you have access to a huge data set which contains ample data on failures rather than encountering failures after release. Hence, the compilation of errors is an important part or early stage of a learning process. Failed today? Fail again tomorrow. You’ll be really strong the days after although it might hurt.

Posted on

AI Collusion

In most applications of AI there is one system of AI, for example a specialized service, that performs in isolation from other services. More powerful systems, however, allow for the combination of AI services. This may be useful in case of integrating services focusing on specialized sensors to gain a more complete impression of the performance of a system. As soon as two and more AI systems become integrated the risk of unwanted or illegal collusion may occur.
Collusion is defined in the realm of economic theory as the secret, undocumented, often illegal, restriction of competition originating from at least two otherwise rival competitors. In the realm of AI collusion has been defined by Motwani et al. (2024) as “teams of communicating generative AI agents solve joint tasks”. The cooperation of agents as well as the sharing of of previously exclusive information increase the risks of violation of rights of privacy or security. The AI related risks consist also in the dilution of responsibility. It becomes more difficult to identify the origin of fraudulent use of data like personal information or contacts. Just imagine using Alexa and Siri talking to each other to develop another integrated service as a simplified example.
The use of steganography techniques, i.e. the secret embedding of code into an AI system or image distribution, can protect authorship as well as open doors for fraudulent applications. The collusion of AI systems will blur legal borders and create multiple new issues to resolve in the construction and implementation of AI agents. New issues of trust in technologies will arise if no common standards and regulations will be defined. We seem to be just at the entry of the new brave world or 1984 in 2024.
(Image: KI MS-Copilot: Three smartphones in form of different robots stand upright on a desk in a circle. Each displays text on a computer image.)

Posted on

AI input

AI is crucially dependent on the input it is built on. This has been already the foundation principle of the powerful search engines like Google that have become to dominate the commercial part of the internet. The crawling of pages on the world wide web and classifying/ranking them with a number of criteria has been the successful business model. The content production was and is done by billions of people across the globe. Open access facilitates the amount of data available.
The business case for AI is not much different. At the 30th anniversary of the “Robots Exclusion Standard” we have to build on these original ideas to rethink our input strategies for AI as well. If there are parts of our input we do not AI to use in its algorithms we have to put up red flags in form of unlisting parts of the information we allow for public access. This is standard routine we might believe, but everything on the cloud might have made it much easier for owners of the cloud space to “crawl” your information, pictures or media files. Some owners of big data collections have decided to sell the access and use to their treasures. AI can then learn from these data.
Restrictions become also clear. More up-to-date information might not be available for AI-treatment. AI might lack the most recent information, if it a kind of breaking news. The strength of AI lies in the size of data input it can handle and treat or recombine. The deficiency of AI is not to know whether the information it uses (is in the data base) is valid or trustworthy. Wrong or outdated input due to a legal change or just-in-time change will be beyond its scope. Therefore, the algorithms have a latent risk involved, i.e. a bias towards the status quo. But the learning algorithms can deal with this and come up with a continued learning or improvement of routines. In such a process it is crucial to have ample feedback on the valid or invalid outcome of the algorithm. Controlling and evaluating outcomes becomes the complementary task for humans as well as AI. Checks and balances like in democratic political systems become more and more important.

Posted on

Sepsis

Sepsis is a major cause of mortality. Therefore, early detection of sepsis is of high importance. Antibiotics constitute a powerful antidote. However, the application of antibiotics without need, i.e. for purely risk reduction in general, has side effects in antibiotics losing their effectiveness later on.
The paper published in The Lancet Digital Health by van der Weijden et al. (2024) reports on the effort to provide an open source access to a calculator of early onset of sepsis (Link). The Neonatal early-onset sepsis calculator developed by Kaiser Permanente builds on the use on the risk carried by mothers like time since membrane rupture, regional infection risks of mothers per 1000 population and the infants presentation at birth. It is important to point out the combination of risks put into the calculator. New systems of artificial intelligence might equally make predictions or recommendations about the application of antibiotics implicitly making use of such a calculator without disclosure.
From a sociological point of view it is interesting to scrutinize the indicators used in the calculation. The approximation of mothers carrying a sepsis risk relies on national, regional or better local indicators. This information is rarely accessible to the public. The choice of a hospital, speed of access to it in case of membrane rupture as well as staffing come into the calculation of an overall risk of sepsis.
It is great to follow the progress of digital health and the increased transparency of critical health decisions at the earliest stages of the life course. Inflammation as a precursor of sepsis should be taken serious at all stages of the life course. (Image calculation based on Kaiser Permanente digital tool Link)

Posted on

AI and Behavior

We start to analyze the impact of AI on our behavior. It is an important question to be aware of not only how we interact with AI (Link), but also what effect the use of AI (disclosed or not) will have on our social behavior. Knowing that AI is used might change our willingness to cooperate or increase or decrease pro-social behavior. The use of AI in form of an algorithm to select job candidates might introduce a specific bias, but it can equally be constructed to favour certain criteria in the selection of candidates. The choice of criteria becomes more important in this process and the process of choosing those criteria.
Next comes the question whether the announcement includes as information that AI will be used in the selection process. This can be interpreted by some that a “more objective” procedure might be applied, whereas other persons interpret this signal as bad sign of an anonymous process and lack of compassion prevalent in the organization focused mostly on efficiency of procedures.  Fabian Dvorak, Regina Stumpf et al. (2024) demonstrate with experimental evidence from various forms of games (prisoner’s dilemma, binary trust game, ultimatum game) that a a whole range of outcomes is negatively affected (trust, cooperation, coordination and fairness). This has serious consequences for society. The social fabric might worsen if AI is widely applied. Even or particularly the undisclosed use of AI already shows up as a lack of trust in the majority of persons in these experiments.
In sum, we are likely to change our behavior if we suspect AI is involved the selection process or content creation. This should be a serious warning to all sorts of content producing media, science, public and private organizations. It feels a bit like with  microplastic or PFAS. At the beginning we did not take it seriously and then before long AI is likely to be everywhere without us knowing or aware of the use. (Image taken on Frankfurt book fair 2017-10!)

Posted on

AI or I

Generative AI receives a lot of attention. One of the main issues is, to study how AI interacts with humans. The hiring decision by managers or an AI algorithm is an interesting application. According to Marie-Pierre Dargnies et al. (2022) the preference for human decisions remains strong despite reasonably unbiased performance of an algorithm. The main issue is with the transparency of the algorithmic decision-making. As a worker or as a hiring manager the preferences continue to sit with the person rather than the AI. It is a worrying outcome, however, that if the rule of gender equality is removed from the algorithm both workers and managers tend to prefer the algorithmic outcome. I interpret this as a latent preference of study participants for gender bias, which could lead them to expect a more favoured outcome in case the AI makes the decision. Knowing what decision-making rules have gone into the hiring algorithm has an impact on all persons involved.
A new managerial competence is to be able to assess tasks carefully, whether you should perform the task yourself or delegate to AI. In this sense the old question: to do the task yourself or to delegate has simply been enlarged by an additional delegation option. The decision-tree goes from (1) To delegate or not to delegate, and (2) if I want/need to delegate, should I delegate to AI or somebody in person (not allowed to use AI).
I opted to use AI for image creation rather than to take a photo myself or by one from a professional photographer. (Image creation: NEUROFLASH AI – Image-Flash 2024-1-26)