Xiyin Tang, Michael Karanicolas, Tyler Emeney, Yan Sun, Nathan Siegel, Nicholas Wilson

Companies and institutions are moving full force to implement artificial intelligence (AI) and its many applications with the potential and promise of making work better, faster and cheaper. However, there are risks of using AI-based systems mindlessly.
Many companies are using AI-based hiring tools to rate job candidates with scores, and the tool taught itself that male candidates were preferable to female candidates. In one case, the system penalized resumes that included the word “women’s,” as in “women’s chess club captain”, and it downgraded graduates of two all-women’s colleges, according to people familiar with the matter. The system was not rating candidates for software developer jobs and other technical posts in a gender-neutral way, and was later scrapped.
Several companies, like HireVue, use AI-based systems to analyze job candidates’ speech and facial expressions in video interviews to reduce reliance on resumes alone. In several instances, the facial expressions of people of Asian descent were misinterpreted as being passive or dull and not suited for dynamic marketing jobs, a clear misunderstanding by the AI systems about cultural artifacts.
Another well-known example is the use of AI in recommending sentences for those who commit crimes. One such tool, COMPAS (Correctional Offender Management Profiling for Alternative Sanctions) uses an AI algorithm to assess potential recidivism risk. In 2016, ProPublica investigated COMPAS and found that black defendants are almost twice as likely as white defendants to be labeled a higher risk but not actually re-offend, however COMPAS tended to make the opposite mistake with whites, meaning that it was more likely to wrongly predict that white people would not commit additional crimes if released compared to black defendants. This level of discrimination was an anathema to U.S. society. So the question is, why did this AI-based system make such a mistake?
Any AI model that relies so heavily on large amounts of data should ensure that the data itself is not compromised. In the case of recidivism, i.e. committing another crime after being released from prison, what COMPAS had was not conviction data on a subsequent crime, but rather the arrest record of people who had already been in prison once. In fact, in most states in the U.S. African Americans are arrested faster on the suspicion of committing a crime. Hence the data that was used to train the COMPAS model was exactly the wrong data to use and simply reflected prejudice in society rather than criminal behavior.
Given the deep learning (and autonomous learning) nature of the state-of-art AI algorithms, it is likely that biases can be difficult to detect until the end. In both examples above, a clear lesson is that the data that goes into training an AI model has to be examined and de-risked very carefully. Data used for training most AI models captures past history rather than present reality. Data on decisions and outcomes replicates human biases, for instance racism and sexism. The algorithms themselves can create new forms of biases if the criteria that the users provide for determining “success” of any kind are flawed. AI models need to be tested adequately and at regular intervals to de-risk them.
The question also remains why the AI development team did not realize that the outputs from their models were biased, and did not reflect the values of society or even the organization using it. If the AI-team had influential women, the hiring and recruitment system would have immediately caught the problem of recommending only men to hire. Similarly, if the COMPAS team had more minorities the outputs would have been challenged at the source. Teams that develop AI-based models and algorithms need to be diverse to ensure that biases that creep in because of a number of reasons, especially “corrupt” data, could be handled or mitigated quickly.
Then there is generative AI, sometimes called large language models (LLMs), which differs from previous forms of AI and analytics because of its ability to generate new content efficiently, often in “unstructured” forms, for example written text or images. One of the best known generative AI tools is ChatGPT, produced by OpenAI and heavily funded by Microsoft.  There has been a large uptick in companies using LLMs in their operations. One of the biggest dangers of using these LLM AI models is that they are prone to “hallucinations”, i.e. answering questions with plausible but untrue assertions.
For example, when David Smerdon, an economist at the University of Queensland, asked ChatGPT “What is the most cited economics paper of all time?”, it answered “A Theory of Economic History” by Douglass North and Robert Thomas, published in the Journal of Economic History in 1969.  Although a plausible answer, the paper does not exist! It was a hallucination of ChatGPT. Smerdon speculates why ChatGPT came to the wrong answer, i.e. hallucinated:  the most cited economics papers often have “theory” and “economic” in them; if an article starts “a theory of economic . . . ” then “ . . . history” is a likely continuation. Douglass North, Nobel laureate, is a heavily cited economic historian, and he wrote a book with Robert Thomas. In other words, the citation is magnificently plausible. What ChatGPT deals with is not truth; it is plausibility.
As the use of AI becomes more ubiquitous, the risks for those using it and  those affected by decisions made by these systems will increase. It is imperative that companies and institutions adopting AI, whether it be current state-of-art AI, generative AI or certainly artificial general intelligence (AGI), should make sure that human intelligence is the master of domains where artificial intelligence is used.
 
The 2023 edition of the Network Readiness Index, dedicated to the theme of trust in technology and the network society, will launch on November 20th with a hybrid event at Saïd Business School, University of Oxford. Register and learn more using this link. 
For more information about the Network Readiness Index, visit https://networkreadinessindex.org/

Professor G. ‘Anand’ Anandalingam is the Ralph J. Tyser Professor of Management Science at the Robert H. Smith School of Business at the University of Maryland. Previously, he was Dean of Imperial College Business School at Imperial College London from August 2013 to July 2016 and Dean of the Smith School of Business at Maryland from 2007 to 2013. Before joining Smith in 2001, Anand was at the University of Pennsylvania for nearly 15 years where he was the National Center Chair of Resource and Technology Management in both the Penn Engineering School and the Wharton School of Business. Anand received his Ph.D. from Harvard University, and his B.A. from Cambridge University, U.K.
The post NRI 2023: The Risks of Unintelligent Adoption of Artificial Intelligence appeared first on Portulans Institute.

By Urbano Reviglio and Danielle Borges

The digital transformation provoked a disruptive effect on the media sector, particularly due to the decrease in advertising revenues. At the same time, it enabled the diversification of media revenue models, including strategies for audience engagement and distribution, opening the possibility for local media providers to create innovative responses to tackle the challenges posed by digitalisation.  

Departing from the data collected by the roster of researchers involved in the project Local Media for Democracy (LM4D), this blog post is the first of a series of three that will explore best practices in the local and community media sector and offer some examples found across the European Union (EU). The LM4D project considers best practices as innovative responses offered by local media outlets to enhance reach and audience, including new forms of work, journalistic products, or services. This first blog post focuses on best practices in the realm of audience and user engagement, in particular on how news organizations, especially local and community media outlets, are making efforts to engage audiences in the harsh competition of the attention economy.  

One of the most popular strategies we came across during the research phase of the LM4D project is the deployment of e-mail newsletters. They have become popular again as a means of disseminating curated news both by legacy print and by newer digital media. Such “renaissance of newsletters” is occurring for various reasons but mainly to avoid being locked into the ecosystems of social media and to regain a gatekeeping power (Hendrickx et al., 2020). E-mail newsletters offer a high degree of targeting, curated content, analysis and commentaries. However, their structures and content can vary significantly, reflecting different editorial visions, objectives or business models (Jack, 2016). And yet, all these usually allow for personalised, carefully curated news experiences for specific niches. They typically do not rely on algorithms to determine what content is shown to subscribers, giving readers a sense of autonomy and control over the content they receive. Also, in an era where the competition for attention is fierce, newsletters can cut through the noise and provide focused, distraction-free content. Indeed, the majority of users (65%) say they enjoy the “convenience” of the format (Newman, 2022).  

Several of the media outlets investigated and interviewed for this blog post have been using newsletters as the main source of engagement with readers. VierNull, for instance, is a local news media from Düsseldorf, Germany, founded in 2021, that publishes news every weekday in the format of an online newsletter, which are sent to subscribers at 5:30 am. A consistent schedule can indeed create a sense of anticipation and routine (a “news habit”) increasing the likelihood that subscribers will open and read the letter. The newsletter contains only one main story and is structured around a personal introduction, a news section and a recommendation section on the explanation about the main article. The idea behind this structure is to show the reader how the work of journalists develops, the reason for choosing a specific topic, the discussion and research around this topic, and what was learned during the writing phase of the article. This is intended to engage and inform the audience on the journalistic process. The outlet is funded mainly by subscriptions and the fact that they do not rely on advertising revenues makes their newsletter even shorter, reassuring their readers that “all news in the letter is important”, as explains Christian Herrendorf, one of the editors. 

Another media outlet effectively using newsletters is InsideStory, from Athens, Greece. The editors apply the principles of slow journalism, which emphasises in-depth research, meticulous fact-checking, and a deliberate, thoughtful presentation of news and stories. They offer one deeply investigated and detailed article per day. The authors take all the time needed to develop their stories and avoid any simplification or sensational interpretation. This approach, the editors argue, is one of the main reasons why InsideStory is being rewarded in terms of subscriptions and visibility. But it is not all. They are also manifestly independent, non-partisan, as well as proactively transparent: fundamental features for high-quality journalism. Overall, this is a successful recipe to regain trust and attention from audiences. And yet, as the editor in chief assures, not only the news production and consumption are “slow”, but also regaining trust is a slow process. Consistency in their mission is therefore crucial, and newsletters are ideal in conveying that.  

Traditional newsletters can also represent a fruitful strategy to engage audiences. A unique example in this regard is Nyomtassteis, a non-profit media outlet from Hungary founded in 2017, which assembles a weekly newsletter that usually contains 6 different topics. The content is not originally produced in their newsrooms but selected by the editorial board. The newsletter is then printed and freely distributed all over the country by volunteers; anyone can print out an A4 sheet of news and put it in their local mailboxes. “Nyomtassteis” means exactly “you print it” in Hungarian. One of the target audiences is the rural population living in small villages with scarce access to internet and low digital skills. Therefore, editors’ choices also take into consideration topics that can be of interest to this specific group and sometimes to specific areas too. In this case, newsletter can also offer a one-to-one connection between the creator and the audience that can build trust and a sense of intimacy, and they can also include discussions, overall fostering a sense of community and connection. This is another significant value added of newsletters which can be leveraged even in online settings.  

Another thriving format encountered during our research is that of podcasts. They can be defined as episodic series of audio files that can be downloaded, subscribed or simply listened (Newman & Gallo, 2019). Unlike other journalism formats using the transmission of voice, such as radio, podcasts are considered a more active process where listeners demonstrate a greater degree of engagement (Berry, 2016). This allows for episodes being developed in a more demanding way, using immersive and narrative storytelling techniques that can be consumed at an ease pace, as a form of “slow media”, which creates a personal connection between the hosts and listeners. Another feature of podcasts is indeed intimacy, strengthened by the use of headphones. However, at the same time, they also create a “sense of collectivity” by allowing recommendations and sharing among users. Moreover, podcasts are transnational, crossing national borders and flowing online across different platform providers (Bonini, 2022). These reasons may explain why this format has become widespread in Europe with its usage ranging from 27% in Belgium to 45% in Spain (Newman et al. 2023). These shared characteristics are attractive to audiences which are seeking a more deliberate, mindful, and meaningful media experience. Indeed, several news media experiment with podcasts. The above-mentioned outlet Viernull, for example, has produced a podcast on true crime. As its editor argues: “the biggest challenge is to be known and recognised in a world with an information overload”.  

Podcasts can help to gain more visibility. Kulturpunkt is another outlet experimenting with podcasts and one of the first in the Croatian media landscape. According to its editor-in-chief, Ivana Pejić, with its direct approach and accessibility to the production phase, podcasts are also perceived “as a huge step towards democratization of communication outside the authoritative structure of the state radio monopoly”. To disseminate their podcasts, Kulturpunkt also employs radio-based community media on a local level to expand reach, and continually researches and tests various formats of mediation for their content. For instance, they introduced a format of collaborative writing for the reviews of contemporary art, and the reports on social and cultural events are enhanced with extensive visuals and audio snippets.  

The experiences of the news media referred in this blog post show that e-mail newsletters and podcasts are effective formats to connecting with audiences despite the different objectives and profiles of the media outlets. Online newsletters have proved to be a powerful strategy to engage audiences, regain their trust and encourage new subscriptions. Their success of these formats shows that there is room to compete with mainstream social media also by leveraging their limitations, such as a lack of control over personalisation, attention-grabbing strategies, and questionable privacy protections. Similarly, podcasts are powerful tools for engaging audiences, particularly younger ones. Publishers’ willingness is indeed to invest more in podcasts and newsletters, as these two formats have proved to be successful in increasing loyalty and in attracting new subscribers (Newman, 2022). In a fiercely competitive media arena, in which attention is scarce and unscrupulous tactics that threaten journalism quality and erode trust in media are widely deployed, these successful formats emphasise the opportunity to offer informative experiences that truly resonate with audience’s needs and, eventually, engage them by openly building trust, not by stealthily capturing their attention. 

This blog post has been written in the context of the LM4D project.

Sources:  

Berry, Richard (2016) Podcasting: Considering the evolution of the medium and its association with the word ‘radio’. The Radio Journal International Studies in Broadcast and Audio Media, 14 (1) 

Bonini, T. (2022) Podcasting as a hybrid cultural form between old and new media. In M. Lindgren and J. Loviglio, Routledge Companion to Radio and Podcast Studies (pp. 19-29), London: Routledge. 

Jack, A. (2016). Editorial email newsletters: the medium is not the only message. Editorial email newsletters: The medium is not the only message. (Oxford paper most cited) 

Hendrickx, J., Donders, K., & Picone, I. (2020). Innovating journalism by going back in time? The curious case of newsletters as a news source in Belgium. Journalistic metamorphosis: Media transformation in the digital age, 57-68. 

Newman, N., & Gallo, N. (2019). News Podcasts and the Opportunities for Publishers. In Digital News Report. Reuters Institute of Journalism. 

Newman, N. (2022) Journalism, Media, and Technology Trends and Predictions. Digital News Project, Reuters Institute for the Study of Journalism, January 2022. Available at https://reutersinstitute.politics.ox.ac.uk/sites/default/files/2023-01/Journalism_media_and_technology_trends_and_predictions_2023.pdf (last accessed 26.10.23). 

Newman, N., Fletcher, R., Eddy, K., Robertson, C. T., & Nielsen, R. K. (2023). Digital News Report 2023. Available at https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2022/email-news-its-contribution-to-engagement-and-monetisation (last accessed 26.10.23). 
The post Newsletters, podcasts and slow media: successful news media strategies to engage audiences in the attention economy   appeared first on Centre for Media Pluralism and Freedom.

Gene aus der MYC-Familie sind für den menschlichen Organismus essenziell.  Nach derzeitigen Erkenntnissen regulieren sie die Expression der meisten zellulären Gene. Eine Fehlsteuerung von MYC-Proteinen trägt wesentlich zur Entstehung vieler Arten von Krebs bei.  Kein Wunder, dass MYC-Proteine im Fokus der Krebsforschung weltweit stehen. Aus Sicht der Wissenschaft könnten sie das ideale Ziel für neue Wirkstoffe im Kampf gegen Krebs sein.
Tatsächlich ist die Bedeutung von MYC für die Entwicklung von Krebszellen seit Langem bekannt. Die Struktur der MYC-Proteine und ihre molekulare Funktion haben es allerdings bisher verhindert, das Protein direkt pharmakologisch anzugreifen. Bei der Suche nach einer Lösung für dieses Problem ist einem Forschungsteam der Julius-Maximilians-Universität Würzburg (JMU) jetzt möglicherweise ein wichtiger Schritt geglückt: Über einen Kooperationspartner von MYC konnte es im Tierversuch die Entstehung und Entwicklung der Krebstumoren deutlich bremsen.
Publikation in „Life Science Alliance”
Beteiligt an der Studie waren zwei Arbeitsgruppen am Lehrstuhl für Biochemie und Molekularbiologie der JMU von Wolf Elmar, Professor für Tumorsystembiologie, und von Dr. Peter Gallant sowie die Gruppe von Thomas Raabe, Professor für Molekulare Genetik. Die Ergebnisse ihrer Arbeit haben die Wissenschaftler jetzt in der Fachzeitschrift „Life Science Alliance“ veröffentlicht.
„Weil es so schwierig ist, MYC-Proteine direkt anzugreifen, haben wir nach Partnern von MYC gesucht und dabei ein Protein namens SPT5 gefunden“, schildert Elmar Wolf die Vorarbeiten zu dieser Studie. SPT5 stellte sich in der Zellkultur als unverzichtbar für die MYC-abhängige Gen-Transkription in menschlichen Krebszellen heraus. Unklar blieb allerdings, wie wichtig die Interaktion von MYC- und SPT5-Proteinen für das Verhalten von normalen Zellen im Körper ist und ob sich über sie die Entwicklung von Krebszellen würde beeinflussen lassen.
Forschung an der Fruchtfliege
Antworten liefert die jetzt veröffentlichte Studie. „Wir haben mit der Fruchtfliege Drosophila melanogaster gearbeitet – einem bekannten und bewährten Modellsystem der tierischen Entwicklung“, erklärt Peter Gallant. Genauso wie Wirbeltiere – und somit auch der Mensch – besitzen Fruchtfliegen ebenfalls MYC- und SPT5-Proteine.
In ihren Experimenten konnten die Wissenschaftler in einem ersten Schritt nachweisen, dass MYC- und SPT5-Proteine auch im Organismus der Fruchtfliege funktionell zusammenarbeiten. So wurde beispielsweise eine moderate Veränderung der MYC- oder der SPT5-Menge von den Fliegen gut toleriert. Veränderte das Team jedoch sowohl MYC- als auch SPT5-Mengen gleichzeitig, traten bei den Tieren deutliche Defekte auf. „Diese Beobachtungen unterstreichen die Wichtigkeit der MYC-SPT5-Interaktion während der normalen Entwicklung des Organismus“, sagt Thomas Raabe.
Drastische Reduktion des Tumorgewebes
Im nächsten Schritt ging das Forschungsteam der Frage nach, welche Rolle SPT5 bei der Entstehung und Entwicklung von Tumoren einnimmt. Zum Einsatz kamen dafür gentechnisch veränderte Fruchtfliegen, die MYC-abhängige Hirntumoren entwickeln. Im Experiment konnten diese Fliegen zwar schlüpfen, starben aber innerhalb von weniger als zehn Tagen, wohingegen die meisten Kontrolltiere nach zwei Monaten noch am Leben waren.
„Wenn wir jedoch bei diesen Exemplaren die Menge an SPT5 in den Hirntumoren experimentell reduzierten, verdreifachte sich ihre Lebenszeit“, schildert Peter Gallant das zentrale Ergebnis der Studie. Dies ging einher mit einer dramatischen Abnahme der Tumormasse, die allerdings nur vorübergehend war. Die Lebenszeit verlängerte sich auch dann, wenn die Wissenschaftlerinnen und Wissenschaftler die SPT5-Menge nicht nur im Gewebe der Hirntumoren reduzierten, sondern im gesamten Organismus der Fliege. Analoge Manipulationen der SPT5-Menge in gesunden Kontrolltieren hatten nur vernachlässigbare Auswirkungen auf die Gehirnstruktur und das Überleben der Tiere.
Nach Aussicht der Würzburger Arbeitsgruppen zeigen diese Resultate, dass SPT5 eine wichtige Rolle bei der Entwicklung von MYC-abhängigen Tumoren spielt. Ihre Experimente lassen auch den Schluss zu, dass eine moderate Reduktion von SPT5 in gesundem Gewebe gut toleriert wird, aber zu einer deutlichen Rückbildung von Tumoren führen kann. Damit erweise sich SPT5 als ein mögliches Zielprotein für die Entwicklung von pharmakologischen Hemmstoffen für die Krebsbekämpfung.
Originalpublikation
Spt5 interacts genetically with Myc and is limiting for brain tumor growth in Drosophila. Julia Hofstetter, Ayoola Ogunleye, André Kutschke, Lisa Marie Buchholz, Elmar Wolf, Thomas Raabe, Peter Gallant. Life Science Alliance Vol 7, Issue 1; doi: 10.26508/lsa.202302130.
Kontakt
Peter Gallant, Lehrstuhl für Biochemie und Molekularbiologie, peter.gallant@uni-wuerzburg.de