Principles and Validations of an Artificial Intelligence-Based Recommender System Suggesting Acceptable Food Changes.

BACKGROUND: Along with the popularity of smartphones, artificial intelligence-based personalized suggestions can be seen as promising ways to change eating habits toward more desirable diets. OBJECTIVES: Two issues raised by such technologies were addressed in this study. The first hypothesis tested is a recommender system based on automatically learning simple association rules between dishes of the same meal that would make it possible to identify plausible substitutions for the consumer. The second hypothesis tested is that for an identical set of dietary-swaps suggestions, the more the user is-or thinks to be-involved in the process of identifying the suggestion, the higher is their probability of accepting the suggestion. METHODS: Three studies are presented in this article, first, we present the principles of an algorithm to mine plausible substitutions from a large food consumption database. Second, we evaluate the plausibility of these automatically mined suggestions through the results of online tests conducted for a group of 255 adult participants. Afterward, we investigated the persuasiveness of 3 suggestion methods of such recommendations in a population of 27 healthy adult volunteers through a custom designed smartphone application. RESULTS: The results firstly indicated that a method based on automatic learning of substitution rules between foods performed relatively well identifying plausible swaps suggestions. Regarding the form that should be used to suggest, we found that when users are involved in selecting the most appropriate recommendation for them, the resulting suggestions were more accepted (OR = 3.168; P < 0.0004). CONCLUSIONS: This work indicates that food recommendation algorithms can gain efficiency by taking into account the consumption context and user engagement in the recommendation process. Further research is warranted to identify nutritionally relevant suggestions.

Measuring implicit associations between food and body stimuli in anorexia nervosa: a Go/No-Go Association Task.

PURPOSE: The present study aimed to explore the implicit associations between food and bodily stimuli in patients with anorexia nervosa (AN) and control subjects (HC). METHODS: A Go/No-Go Association Task was administrated to 55 participants (28 AN and 27 HC), using food stimuli (low-calorie food vs. high-calorie food) and body stimuli (underweight vs. overweight bodies). RESULTS: We evidenced an implicit association between food and body stimuli in the AN group, whereas the HC group only showed a tendency. AN and HC groups also exhibited different categorization strategies: the AN group tended to categorize stimuli as low-calorie foods and underweight bodies less than the HC group, and they tended to categorize stimuli as high-calorie foods and overweight bodies more than the HC group. CONCLUSION: The present study revealed for the first time specificities of the AN population's implicit association between food and body stimuli in terms of association strength and categorization strategy. Furthermore, the results suggest that combining implicit methodologies with other methods could contribute to a better characterization of the physiopathology of AN. LEVEL OF EVIDENCE: Level I, experimental study.

Diet Protein Content and Individual Phenotype Affect Food Intake and Protein Appetence in Rats.

BACKGROUND: A low-protein diet can induce compensatory intake of excess energy. This must be better evaluated to anticipate the obesogenic risk that may result from the dietary recommendations for reducing animal protein consumption. OBJECTIVES: We aimed to further characterize the behavioral and physiological responses to a reduction in dietary protein and to identify the determinants of protein appetite. METHODS: Thirty-two male Wistar rats [4 wk old, (mean ± SEM) 135 ± 32 g body weight] were fed a low-protein (LP; 6% energy value) or normal-protein (NP; 20%) diet for 8 wk. Food intake and body mass were measured during the entire intervention. During self-selection sessions after 4 wk of experimental diets, we evaluated rat food preference between LP, NP, or high-protein (HP; 55%) pellets. At the end of the experiment, we assessed their hedonic response [ultrasonic vocalizations (USVs)] and c-Fos neuronal activation in the olfactory tubercle and nucleus accumbens (NAcc) associated with an LP or HP meal. RESULTS: Rats fed an LP diet had greater food intake (24%), body weight (5%), and visceral adiposity (30%) than NP rats. All LP rats and half of the NP rats showed a nearly exclusive preference for HP pellets during self-selection sessions, whereas the other half of the NP rats showed no preference. This suggests that the appetite for proteins is driven not only by a low protein status but also by individual traits in NP rats. LP or HP meal induced similar USV emission and similar neuronal activation in the NAcc in feed-deprived LP and NP rats, showing no specific response linked to protein appetite. CONCLUSIONS: Protein appetite in rats is driven by low protein status or individual preferences in rats receiving adequate protein amounts. This must be considered and further analyzed, in the context of current recommendations for protein intake reduction.

Social modeling of food choices in real life conditions concerns specific food categories.

The social context of eating has a profound effect on consumption choices. Social modeling, that involves using others' behavior as a guide for appropriate consumption, has been well documented for food intake, but less is known about social modeling of food choices. Moreover, social modeling has mainly been studied in laboratory settings. We conducted an observational study in a self-service canteen to examine whether the food choices of an individual were influenced by the choice of the person ahead in the queue. We recorded food choices of 546 individuals (333 men and 211 women) and those of the person in front of them in the queue along a linear buffet. Starters were sub-categorized into salads, mixed starters (e.g. avocado shrimp mayonnaise), and cold meat starters, and desserts were sub-categorized into fruits, dairy products and pastries. There was a significantly higher probability of taking a starter in general (OR = 1.65, IC = 1.06-2.57, p = 0.03), a salad (OR = 1.78, CI = 1.08-2.93, p = 0.02), a mixed starter (OR = 2.98, CI = 1.42-6.05, p < 0.01), but not a cold meat, if the person ahead in the queue also took one compared to when the person ahead did not take one. No significant modelling was found for desserts which may be because almost all participants took a dessert. These results highlight that social modeling influences food choices, and that this phenomenon can be observed in a real life setting. These data also suggest that some food categories, such as starters, could be more susceptible to social modeling than are others. Finally, we observed modeling both between familiar and unfamiliar participants, which suggests that social norms could be used to promote healthier eating in a range of settings including friendship groups.

Self-declared attitudes and beliefs regarding protein sources are a good prediction of the degree of transition to a low-meat diet in France.

Meat consumption in Western countries is declining and, while the proportion of strict vegetarians remains low, intermediate diets such as flexitarianism have been developing in recent years. Our objectives were to identify the different levels of transition towards low-meat diets, characterize how these diets differ in terms of food intake, and identify whether attitudes and beliefs can explain these degrees of transition. In a representative survey of the French adult population conducted in 2018 (n = 2055), participants declared whether they followed a particular diet and completed a food frequency questionnaire on 29 food sources of protein and a questionnaire on their attitudes and beliefs regarding protein sources. We identified four dietary types based on these declarative data: vegetarians, flexitarians, pro-flexitarians and omnivores. The theory of planned behavior was used to predict meat intake and intentions to reduce meat intake. The sample contained 2.5% vegetarians, 6.3% flexitarians, 18.2% pro-flexitarians and 72.9% omnivores. The diet groups displayed specific dietary profiles and attitudinal scores. Compared with omnivores, pro-flexitarians consumed less red meat, more vegetables and legumes and were much more in agreement about the environmental impacts of meat. Compared with pro-flexitarians, flexitarians consumed less red meat and processed meat, and agreed much more about the health impacts of meat. Finally, versus flexitarians, vegetarians consumed almost no meat but far more legumes, nuts and seeds, and were much more sensitive to animal welfare issues. Attitudes, social norms and perceived behavioral control (PBC) predicted intentions to reduce meat consumption but attitude was the most important predictor. Intentions and PBC were both predictive of meat consumption. The dietary type related to the level of meat intake could be predicted by self-declared attitudes and beliefs regarding protein sources.

Differences in BOLD responses in brain reward network reflect the tendency to assimilate a surprising flavor stimulus to an expected stimulus.

External information can modify the subjective value of a tasted stimulus, but little is known about neural mechanisms underlying these behavioral modifications. This study used flavored drinks to produce variable degrees of discrepancy between expected and received flavor. During a learning session, 43 healthy young men learned 4 symbol-flavor associations. In a separate session, associations were presented again during an fMRI scan, but half of the trials introduced discrepancy with previously learned associations. Liking ratings of drinks were collected and were analyzed using a linear model to define the degree to which discrepant symbols affected liking ratings of the subjects during the fMRI session. Based on these results, a GLM analysis of fMRI data was conducted to determine neural correlates of observed behavior. Groups of subjects were composed based on their behavior in response to discrepant symbols, and comparison of brain activity between groups showed that activation in the PCC and the caudate nucleus was more potent in those subjects in which liking was not affected by discrepant symbols. These activations were not found in subjects who assimilated unexpected flavors to flavors preceeded by discrepant symbols. Instead, these subjects showed differences in the activity in the parietal operculum. The activity of reward network appears to be related to assimilation of received flavor to expected flavor in response to symbol-flavor discrepancy.

Expected satiation alone does not predict actual intake of desserts.

The degree to which consumers expect foods to satisfy hunger, referred to as expected satiation, has been reported to predict food intake. Yet this relationship has not been established precisely, at a quantitative level. We sought to explore this relationship in detail by determining whether expected satiation predicts the actual intake of semi-solid desserts. Two separate experiments were performed: the first used variations of a given food (eight apple purées), while the second involved a panel of different foods within a given category (eight desserts). Both experiments studied the consumption of two products assigned to volunteers based on their individual liking and expected satiation ratings, given ad libitum at the end of a standardised meal. A linear model was used to find predictors of food intake and included expected satiation scores, palatability scores, BMI, age, sex, TFEQ-R, TFEQ-D, water consumption during the meal, reported frequency of eating desserts, and reported frequency of consuming tested products as explanatory variables. Expected satiation was a significant predictor of actual food intake in both experiments (apple purée: F(1,97) = 18.60, P < .001; desserts: F(1,106) = 9.05, P < .01), along with other parameters such as product palatability and the volunteers' age, sex and food restriction (variation explained by the model/expected satiation in the experiments: 57%/23% and 36%/17%, respectively). However, we found a significant gap between expected and actual consumption of desserts, on group and on individual level. Our results confirm the importance of expected satiation as a predictor of subsequent food intake, but highlight the need to study individual consumption behaviour and preferences in order to fully understand the role of expected satiation.

Assimilation and contrast are on the same scale of food anticipated-experienced pleasure divergence.

Consumption of a product is preceded by an anticipation of its qualities by the consumer, which can itself modify the consumption experience. Improved knowledge of anticipation would allow better manipulation of it, for example to enhance the acceptance of healthier foods. According to the Assimilation-Contrast theory, the size of anticipation-reality divergence determines how anticipation influences consumers' satisfaction. For small divergences, experienced pleasure is the same as the anticipated pleasure (Assimilation); for large ones, the effect of surprise provokes an even larger discordance with that which was anticipated (Contrast). Few studies have attempted to observe both effects simultaneously, or to consider the anticipation-reality divergence quantitatively rather than qualitatively; these were the study's objectives. A range of 10 flavored drinks was developed to vary progressively in intensity. Ninety healthy young men consumed samples during two separate sessions. In session 1, hedonic and sensory scores of all drinks were recorded during blind tasting. In session 2, three drinks were chosen as references for taste intensity, and associated with neutral symbols that served as labels. Subjects then consumed 36 drink samples, each one bearing a label. For half of the samples drinks did not correspond to labels, creating a range of anticipation-reality divergence. By predicting session 2 scores using linear modeling with session 1 blind ratings as input, it was confirmed that both Assimilation and Contrast effects on hedonic ratings were present (Assimilation (t(89) = 5.645, p < 0.0001) and Contrast (t(89) = 3.186, p = 0.002 or t(89) = 2.494, p = 0.015, depending on the drink-label combination)). This study was the first to position Assimilation and Contrast within a quantitative context using controlled divergence variation rather than products from distinct categories.

Environmental enrichment alters splenic immune cell composition and enhances secondary influenza vaccine responses in mice.

Chronic stress has deleterious effects on immune function, which can lead to adverse health outcomes. However, studies investigating the impact of stress reduction interventions on immunity in clinical research have yielded divergent results, potentially stemming from differences in study design and genetic heterogeneity, among other clinical research challenges. To test the hypothesis that reducing glucocorticoid levels enhances certain immune functions, we administered influenza vaccine once (prime) or twice (boost) to mice housed in either standard control caging or environmental enrichment (EE) caging. We have shown that this approach reduces mouse corticosterone production. Compared with controls, EE mice had significantly lower levels of fecal corticosterone metabolites (FCMs) and increased splenic B and T lymphocyte numbers. Corticosterone levels were negatively associated with the numbers of CD19(+) (r(2) = 0.43, p = 0.0017), CD4(+) (r(2) = 0.28, p = 0.0154) and CD8(+) cells (r(2) = 0.20, p = 0.0503). Vaccinated mice showed nonsignificant differences in immunoglobulin G (IgG) titer between caging groups, although EE mice tended to exhibit larger increases in titer from prime to boost than controls; the interaction between the caging group (control versus EE) and vaccine group (prime versus boost) showed a strong statistical trend (cage-group*vaccine-group, F = 4.27, p = 0.0555), suggesting that there may be distinct effects of EE caging on primary versus secondary IgG vaccine responses. Vaccine-stimulated splenocytes from boosted EE mice had a significantly greater frequency of interleukin 5 (IL-5)-secreting cells than boosted controls (mean difference 7.7, IL-5 spot-forming units/10(6) splenocytes, 95% confidence interval 0.24-135.1, p = 0.0493) and showed a greater increase in the frequency of IL-5-secreting cells from prime to boost. Our results suggest that corticosterone reduction via EE caging was associated with enhanced secondary vaccine responses, but had little effect on primary responses in mice. These findings help identify differences in primary and secondary vaccine responses in relationship to stress mediators that may be relevant in clinical studies.

The effect of social norms on vegetarian choices is moderated by intentions to follow a vegetarian diet in the future: Evidence from a laboratory and field study.

Social norms could be a tool in dietary transition toward more sustainable diets, but the results of social norms interventions aimed at encouraging the selection of plant-based foods to date have been inconsistent. One reason for this might be because there are important moderating factors that have yet to be investigated. Here we examine social modeling of vegetarian food choices and test whether modeling is dependent upon individual intentions to follow a vegetarian diet in the future in two different settings. In a laboratory study of 37 women, participants with low intentions to become a vegetarian consumed fewer plant-based foods in the presence of a vegetarian confederate, compared to eating alone. In an observational study of 1,037 patrons of a workplace restaurant, participants with a higher score of on vegetarian intentions had a greater likelihood of taking a vegetarian main course or starter, and a vegetarian social norm was associated with a greater likelihood of a vegetarian choice for the main course but not for the starter. These data suggest that participants with low intentions to follow a vegetarian diet may exhibit reactance against an explicit vegetarian norm in an unfamiliar context (as in Study 1) but that general norm following regardless of dietary intentions be more likely when it is conveyed implicitly in a familiar context (as in Study 2).

Dopamine Modulates the Processing of Food Odour in the Ventral Striatum.

Food odour is a potent stimulus of food intake. Odour coding in the brain occurs in synergy or competition with other sensory information and internal signals. For eliciting feeding behaviour, food odour coding has to gain signification through enrichment with additional labelling in the brain. Since the ventral striatum, at the crossroads of olfactory and reward pathways, receives a rich dopaminergic innervation, we hypothesized that dopamine plays a role in food odour information processing in the ventral striatum. Using single neurones recordings in anesthetised rats, we show that some ventral striatum neurones respond to food odour. This neuronal network displays a variety of responses (excitation, inhibition, rhythmic activity in phase with respiration). The localization of recorded neurones in a 3-dimensional brain model suggests the spatial segregation of this food-odour responsive population. Using local field potentials recordings, we found that the neural population response to food odour was characterized by an increase of power in the beta-band frequency. This response was modulated by dopamine, as evidenced by its depression following administration of the dopaminergic D1 and D2 antagonists SCH23390 and raclopride. Our results suggest that dopamine improves food odour processing in the ventral striatum.

Associations between Perceived Social Eating Norms and Initiation and Maintenance of Changes in Dietary Habits during the First COVID-19 Lockdown in France.

Changes in dietary habits of the French population have been reported during the national lockdown that was enforced due to the COVID-19 pandemic. This study investigated whether perceived social eating norms were associated with the initiation and maintenance of dietary changes that took place as a result of lockdown. An online study collected information on (1) changes in consumption implemented during the lockdown and the maintenance of these changes, and (2) perceptions about changes in consumption implemented during lockdown by household members, relatives out of home, and the general population. The changes in consumption were classified as foods to increase or to decrease according to French national recommendations. The perception of changes to dietary habits by household members and relatives out of home was related to the changes made by individuals for each of the food categories (all p < 0.05) but not to the perception of changes made the general population. Increased consumption of foods to increase was more likely to be maintained when there was a positive perception of the changes made by household members (p = 0.03). These results highlight the influence of the perception of social eating norms, especially by household members and relatives, on the implementation of dietary changes during lockdown and suggest that social eating norms can have a lasting influence.

The Willingness to Modify Portion Sizes or Eat New Protein Foods Largely Depends on the Dietary Pattern of Protein Intake.

Promoting a more balanced animal/plant dietary protein ratio by changing portion sizes or introducing new foods is a promising means to improve diet quality, but little is known about the willingness of individuals to adopt such changes. Our objective was to assess the willingness to adopt dietary changes by these means. In a French cross-sectional study in 2018 (n = 2055), we analyzed the association between the willingness to eat smaller or larger portions or to introduce non-consumed protein foods and the current dietary patterns of individuals and their socio-demographic characteristics. These modifications had previously been identified as improving the nutrient adequacy of diets. Participants were more willing to eat smaller portion sizes than to introduce new foods and to eat larger portion sizes. The willingness for any modification varied depending on the food groups concerned. Participants were also more willing to eat larger portions and less willing to eat smaller portions when they were the most frequent consumers of the foods concerned. Participants were more willing to eat a new food if it was consumed in large quantities by individuals with a similar dietary pattern. This study underlines the importance of accounting for individual food habits when issuing nutritional recommendations.